Methods of diagnosing and treating acute respiratory infections

Abstract

The present invention provides methods for determining a cause of an acute respiratory infection in a subject and methods of treating a subject based on the determination, as well as systems useful for making the determination using a biological sample from a subject.

Description

[0001] This application is a continuation-in-part of U.S. Patent Application Serial No. 15 / 191, 1 19, filed June 30, 2016, which claims priority to U.S. Provisional Patent Application Serial No. 62 / 187,683, filed July 1, 2015, and U.S. Provisional Patent Application Serial No. 62 / 257,406, filed November 19, 2015, the disclosures of each of which are incorporated herein by reference in their entirety.

[0002] Related Applications

[0003] This application claims the benefit of U.S. Provisional Patent Application Serial No. 62 / 187,683, filed July 1, 2015, and U.S. Provisional Patent Application Serial No. 62 / 257,406, filed November 19, 2015, the disclosures of each of which are incorporated herein by reference in their entirety.

[0004] FEDERALLY SPONSORED RESEARCH

[0005] This application was made with federal funds from the National Institutes of Health (NIH) under grant numbers U01 AI066569, P20 RR016480, and HHSN266200400064C, and from the Defense Advanced Research Projects Agency (DARPA) under grant numbers N66001-07-C-2024 and N66001-09-C-2082. The United States Government has certain rights in the application. BACKGROUND

[0006] Acute respiratory infections are common in acute care settings and cause significant mortality, morbidity, and economic loss worldwide. Respiratory tract infections, or acute respiratory infections (ARI), caused 3.2 million deaths worldwide and 164 million disability-adjusted life years of lost life in 2011, more than any other cause (WHO, 2013a, 2013b). In 2012, lower respiratory tract infections were the fourth leading cause of death worldwide, and were the leading cause of death in low- and middle-income countries where less supportive care is available (WHO News Release, 22 August 2014). These diseases are also a problem in developed countries. In the United States, in 2010, the Centers for Disease Control (CDC) determined that pneumonia and influenza alone caused 15.1 deaths per 100,000 people in the U.S. population. Older adults and children under 5 years of age are particularly vulnerable to bad outcomes due to ARIs. For example, in 2010, pneumonia accounted for 18.3% of all deaths worldwide in children 5 years of age or younger, or almost 1.4 million deaths.

[0007] Pneumonia and other lower respiratory tract infections can be due to many different pathogens, which are primarily viral, bacterial, or less commonly, fungal. Among viral pathogens, influenza is the most notable one, based on the number of individuals infected, the variable severity from season to season, and the often-present concern about new strains that cause much higher morbidity and mortality (e.g., avian flu). However, among viral pathogens, influenza is only one of many that cause significant human disease. Respiratory Syncytial Virus (RSV) is the leading cause of hospitalization in children during winter months in developed countries. Worldwide, in 2005, about 33 million new cases of RSV infection were reported in children under 5 years of age, and 3.4 million were severe enough to require hospitalization. It is estimated that between 66,000 and 199,000 children die each year from this viral infection alone. And, in the U.S. alone, about 10,000 people over 65 years of age die each year with RSV infection. In addition to known viral pathogens, history shows that new and emerging infections can appear at any time, spread globally within days or weeks. Recent examples include SARS - a coronavirus - when it emerged in 2003-2004, with a 10% mortality rate. More recently, the Middle East Respiratory Syndrome (MERS) coronavirus continues to brew in the Middle East and is associated with a 30% mortality rate. Both of these infections exhibit respiratory symptoms and can not be distinguished from any other ARI at the outset.

[0008] Although viral infections cause the majority of ARIs, bacterial etiologies are also important, especially in the case of lower respiratory tract infections. The specific causes of bacterial ARIs vary geographically and according to the clinical setting, but include Streptococcus pneumoniae, Staphylococcus aureus, Haemophilus influenzae, Chlamydia pneumoniae, Mycoplasma pneumoniae, Klebsiella pneumoniae, Escherichia coli, and Pseudomonas aeruginosa. Identification of these pathogens relies on their growth in culture, which typically takes several days and has limited sensitivity for the detection of infectious agents. Obtaining a sufficient sample for testing is problematic: in a study of 1669 patients with community-acquired pneumonia, only 14% of patients could provide a "good quality" sputum sample that yielded a positive culture (Garcia-Vazquez et al., 2004). Clinicians are aware of the limitations of these tests, which introduce uncertainty, and as a result, antibacterial therapy is often prescribed without any confirmation of bacterial infection.

[0009] The ability to rapidly diagnose the cause of ARI is an urgent global problem with far-reaching implications at multiple levels: optimizing treatment for individual patients; identifying and tracking outbreaks for epidemiological surveillance; and guiding appropriate use of antimicrobials to stem the rising tide of antimicrobial resistance. It is widely accepted that early and appropriate antimicrobial therapy improves outcomes in patients with serious infections. This has in part contributed to the overuse of antimicrobial therapy. Up to 73% of ambulatory patients with acute respiratory illness are prescribed antibiotics, accounting for approximately 40% of all antibiotic prescriptions for adults in this setting (Cantrell et al., 2003, Clin. Ther. Jan; 24(1): 170-82). However, it is estimated that only a small fraction of these patients require antibacterial therapy (Cantrell et al., 2003, Clin. Ther. Jan; 24(1): 170-82). Similar trends are observed in emergency departments. Even when the presence of a viral pathogen has been microbiologically confirmed, the possibility of a concurrent bacterial infection is not excluded. As a result, antibacterial agents are often prescribed "just in case". This spiraling empiricism contributes to the rising tide of antimicrobial resistance (Gould, 2009; Kim & Gallis, 1989), which in itself is associated with higher mortality, length of hospital stay, and health care costs (Cosgrove 2006, Clin. Infect. Dis., Jan 15; 42 Suppl 2:S82-9). Furthermore, inappropriate use of antibiotics can lead to drug-related side effects and other complications, such as Clostridium difficile-associated diarrhea (Zaas et al., 2014).

[0010] Acute respiratory infections are often characterized by non-specific symptoms (such as fever or cough) that are common to many different diseases, including those not caused by infection. Existing diagnostics for ARI are deficient in many ways. Conventional microbiology tests are limited by low sensitivity and specificity, slow turn-around time, or complexity of the assay (Zaas et al. 2014, Trends Mol Med 20(10):579-88). One limitation of current assays to detect specific viral pathogens (e.g., multiplex PCR-based assays) is the inability to detect emerging or prevalent viral strains. Influenza epidemics occur when a new virus is transmitted in a population that has no natural resistance to it. Influenza epidemics are often devastating. For example, in 1918-1919, the Spanish flu infected about 20% to 40% of the world's population and killed about 50 million people; in 1957-1958, the Asian flu killed about 2 million people; in 1968-1969, influenza killed about 1 million people; and in 2009-2010, the Centers for Disease Control estimates that between about 43 and 89 million people were infected with swine flu, resulting in 8,870 to 18,300 related deaths. The emergence of these new strains challenges existing diagnostics that were not designed to detect them. This was particularly evident during the 2009 influenza epidemic, where confirmation of infection took days and was performed only at specialized testing centers like state health departments or the CDC (Kumar & Henrickson 2012, Clin Microbiol Rev 25(2):344-61). The current outbreak of Ebola virus disease in West Africa is raising similar challenges. Furthermore, it is expected from time to time that we will continue to face this problem, as future outbreaks of infectious disease are inevitable.

[0011] A further limitation of diagnostics using the paradigm of testing for a specific virus or bacteria is that even if a pathogenic microbe can be detected, this does not prove that the patient's symptoms are due to the detected pathogen. Microbes can exist as part of an individual's normal flora, which is called colonization, or they can be detected due to contamination of the sample being tested (e.g., a nasal swab or wash). While recently approved multiplex PCR assays, including those that detect both viruses and bacteria, provide high sensitivity, these tests cannot distinguish between the asymptomatic carriage of a virus and a true infection. For example, there is a high proportion of asymptomatic viral shedding in ARI, especially in children (Jansen et al. 2011, J Clin Microbiol 49(7):2631-2636). Similarly, even if one pathogen is detected, the disease can be attributed to a second pathogen for which a test is not available or was not performed.

[0012] Reports have described host gene expression profiles that distinguish viral ARI from healthy controls (Huang et al. 2011 PLoS Genetics 7(8):e1002234; Mejias et al. 2013; Thach et al. 2005 Genes and Immunity 6:588-595; Woods et al. 2013; A. K. Zaas et al. 2013; A. K. Zaas et al. 2009). However, few of these distinguish viral from bacterial ARI, which is a clinically more meaningful distinction than the detection of viral infection versus health or bacterial infection versus health (Hu, Yu, Crosby, & Storch, 2013; Parnell et al. 2012; Ramilo et al. 2007).

[0013] Current diagnostic methods are therefore limited in their ability to distinguish between bacterial and viral infections and the appearance of symptoms of non-infectious causes, or to discriminate between combined infections with bacteria and viruses.

[0014] SUMMARY

[0015] The present invention provides, in part, molecular diagnostic tests that overcome many of the limitations of current methods for determining the etiology of respiratory system symptoms. The tests detect the host's response to one or more infectious agents by measuring and analyzing co-expressed genes or signatures. These gene expression signatures can be measured in blood samples in humans or animals presenting symptoms consistent with acute respiratory infection or in humans or animals at risk of developing (prior to symptom onset) an acute respiratory infection (during an epidemic or localized outbreak of disease). Measurement of the host response as taught herein distinguishes between bacterial ARI, viral ARI, and non-infectious etiologies, and can also detect ARIs caused by combined infections with bacteria and viruses.

[0016] This multi-element test is performed with unprecedented accuracy and clinical applicability, allowing health care providers to utilize the host's (subject's or patient's) response to reliably determine the nature of the infectious agent to the level of pathogen species, or to rule out infectious causes of symptoms in individual patients presenting symptoms that are not specific per se. In some embodiments, the result is agnostic as to the species of respiratory virus or bacteria (i.e., in distinguishing between viral or bacterial, the specific genus or species of virus or bacteria cannot be distinguished). This provides an advantage over current tests that include probes or reagents for specific pathogens and are therefore limited to detecting only those specific pathogens.

[0017] One aspect of the present invention provides a method for determining whether an acute respiratory symptom in a subject is of bacterial origin, viral origin, or non-infectious origin, the method comprising, consisting of, or consisting essentially of: (a) obtaining a biological sample from the subject; (b) determining a gene expression profile of the subject from the biological sample by evaluating the expression levels of a predefined set of genes referred to as signatures; (c) normalizing the gene expression levels to a technology (i.e., platform) used to make the measurements to generate normalized values; (d) inputting the normalized values into a bacterial classifier, a viral classifier, and / or a non-infectious disease classifier having predefined weighting values (coefficients) for each of the genes in each signature; (e) comparing the output of the classifiers to predefined threshold, cutoff, or range of values indicative of likelihood of infection; and (f) using the output to determine whether the patient providing the sample has an infection of bacterial origin, viral origin, or a non-infectious disease, or some combination of these conditions.

[0018] Another aspect of the application provides a method for determining whether an acute respiratory infection (ARI) in a subject is of bacterial origin, viral origin, or non-infectious origin, the method comprising, consisting of, or consisting essentially of: (a) obtaining a biological sample from the subject; (b) determining a gene expression profile of the subject from the biological sample by evaluating the expression levels of a predefined plurality of sets of genes; (c) normalizing the gene expression levels to a technology (i.e. platform) used to make the measurements to generate normalized values; (d) inputting the normalized values into a classifier having a predefined weighting value for each of the genes in each identifying signature; e) comparing the output of the classifier to a predefined threshold, cutoff, or range of values indicative of the likelihood of infection; (f) repeating step (d) using only the viral and non-infectious classifiers if the sample is negative for bacteria; and (g) classifying the sample as viral etiology or non-infectious disease.

[0019] Another aspect of the application provides a method for determining whether an acute respiratory infection (ARI) in a subject is of bacterial origin, viral origin, or non-infectious origin, the method comprising, consisting of, or consisting essentially of: (a) obtaining a biological sample from the subject; (b) determining a gene expression profile of the subject from the biological sample by evaluating the expression levels of a predefined plurality of sets of genes; (c) normalizing the gene expression levels to a technology (i.e. platform) used to make the measurements to generate normalized values; (d) inputting the normalized values into a classifier having a predefined weighting value for each of the genes in each identifying signature; e) comparing the output of the classifier to a predefined threshold, cutoff, or range of values indicative of the likelihood of infection; (f) repeating step (d) using only the bacterial and non-infectious classifiers if the sample is negative for viruses; and (g) classifying the sample as bacterial etiology or non-infectious disease.

[0020] Another aspect of the application provides a method for determining whether an acute respiratory infection (ARI) in a subject is of bacterial origin, viral origin, or non-infectious origin, the method comprising, consisting of, or consisting essentially of: (a) obtaining a biological sample from the subject; (b) determining a gene expression profile of the subject from the biological sample by evaluating the expression level of a predefined plurality of sets of genes; (c) normalizing the gene expression level to the technology (i.e. platform) used to make the measurement to generate a normalized value; (d) inputting the normalized value into a classifier having a predefined weighting value for each of the genes in each signature; (e) comparing the output of the classifier to a predefined threshold, cutoff, or range of values indicative of the likelihood of infection; (f) repeating step (d) using only the viral and bacterial classifiers if the sample is negative for non-infectious disease; and (g) classifying the sample as viral or bacterial etiology.

[0021] Yet another aspect of the application provides a method of treating an acute respiratory infection (ARI) in a subject whose etiology is unknown, the method comprising, consisting of, or consisting essentially of: (a) obtaining a biological sample from the subject; (b) determining a gene expression profile of the subject from the biological sample by evaluating the expression level of a predefined plurality of sets of genes (e.g., one, two or more signatures); (c) normalizing the gene expression level to the technology (i.e. platform) used to make the measurement to generate a normalized value; (d) inputting the normalized value into a bacterial, viral, and non-infectious disease classifier having a predefined weighting value for each of the genes in each signature; (e) comparing the output of the classifier to a predefined threshold, cutoff, or range of values indicative of the likelihood of infection; (f) classifying the sample as bacterial, viral, or non-infectious disease etiology; and (g) administering to the subject an appropriate treatment regimen as determined by step (e). In some embodiments, step (g) comprises administering an antibacterial treatment when the etiology of the ARI is determined to be bacterial. In other embodiments, step (g) comprises administering an antiviral treatment when the etiology of the ARI is determined to be viral.

[0022] Another aspect is a method of monitoring a response to a vaccine or a drug in a subject having or at risk of an acute respiratory disease selected from bacterial, viral, and / or non-infectious, the method comprising determining a host response in the subject by a method as taught herein. In some embodiments, the drug is an antibacterial or antiviral drug.

[0023] In some embodiments of this aspect, the method further comprises generating a report that assigns a score to the subject indicative of a probability of an etiology of the ARI.

[0024] Also provided is a system for determining an etiology of an acute respiratory illness in a subject selected from a bacterial, viral, and / or non-infectious origin, the system comprising one or more of the following (including combinations thereof): at least one processor; a sample input circuit configured to receive a biological sample from the subject; a sample analysis circuit coupled with the at least one processor and configured to determine gene expression levels of the biological sample; an input / output circuit coupled with the at least one processor; a storage circuit coupled with the at least one processor and configured to store data, parameters, and / or classifiers; and a memory coupled with the processor and comprising computer readable program code embodied in the memory that, when executed by the at least one processor, causes at least one processor to perform operations comprising: controlling / performing, by the sample analysis circuit, measurement of gene expression levels of a predefined set of genes (i.e., identifying markers) in the biological sample; normalizing the gene expression levels to generate normalized gene expression values; retrieving, from the storage circuit, a bacterial acute respiratory infection (ARI) classifier, a viral ARI classifier, and a non-infectious disease classifier, the one or more classifiers comprising predefined weighting values (i.e., coefficients) for each of the genes of the predefined set of genes; inputting the normalized gene expression values into one or more acute respiratory illness classifiers selected from the bacterial acute respiratory infection (ARI) classifier, the viral ARI classifier, and the non-infectious disease classifier; calculating, based on the one or more classifiers, a probability of an etiology for one or more of a bacterial ARI, a viral ARI, and a non-infectious disease; and controlling output, by the input / output circuit, a determination of whether the acute respiratory illness in the subject is of a bacterial origin, a viral origin, a non-infectious origin, or some combination thereof.

[0025] In some embodiments, the system comprises computer readable code to convert quantitative or semi-quantitative gene expression detection into a cumulative score or probability of an etiology of the ARI.

[0026] In some embodiments, the system comprises an array platform, a thermal cycler platform (e.g., multiplex and / or real-time PCR platform), a hybridization and multi-signal encoded (e.g., fluorescent) detector platform, a nucleic acid mass spectrometry platform, a nucleic acid sequencing platform, or combinations thereof.

[0027] In some embodiments of this aspect, the predefined plurality of sets of genes comprises at least three genetic markers.

[0028] In some embodiments of this aspect, the biological sample comprises a sample selected from the group consisting of peripheral blood, sputum, nasopharyngeal swab, nasopharyngeal wash, bronchoalveolar lavage, endotracheal aspirate, and combinations thereof.

[0029] In some embodiments of this aspect, the bacterial classifier comprises the expression level of 5, 10, 20, 30, or 50, to 80, 100, 150, or 200 of the genes (measurable, e.g., having oligonucleotide probes homologous to the genes or gene transcripts) listed as part of the bacterial classifier in Table 1, Table 2, Table 9, Table 10, and / or Table 12. In some embodiments, the viral classifier comprises the expression level of 5, 10, 20, 30, or 50, to 80, 100, 150, or 200 of the genes (measurable, e.g., having oligonucleotide probes homologous to the genes or gene transcripts) listed as part of the viral classifier in Table 1, Table 2, Table 9, Table 10, and / or Table 12. In some embodiments, the non-infectious disease classifier comprises the expression level of 5, 10, 20, 30, or 50, to 80, 100, 150, or 200 of the genes (measurable, e.g., having oligonucleotide probes homologous to the genes or gene transcripts) listed as part of the non-infectious disease classifier in Table 1, Table 2, Table 9, Table 10, and / or Table 12.

[0030] Also provided is a kit for determining the etiology of an acute respiratory infection (ARI) in a subject, the kit comprising, consisting of, or consisting essentially of (a) means for extracting mRNA from a biological sample; (b) means for generating one or more arrays consisting of a plurality of synthetic oligonucleotides having regions homologous to 5, 10, 20, 30, or 50, to 80, 100, 150, or 200 of the transcripts from the genes of Table 1, Table 2, Table 9, Table 10, and / or Table 12; and (c) instructions for use.

[0031] Another aspect of the application provides a method of using the kit to assess the acute respiratory infection (ARI) classifier, the method comprising, consisting of, or consisting essentially of: (a) generating one or more arrays consisting of a plurality of synthetic oligonucleotides having regions of homology to 5, 10, 20, 30, or 50 to 80, 100, 150, or 200 of the genes from Table 1, Table 2, Table 9, Table 10, and / or Table 12; (b) adding oligonucleotides of regions of homology to normalizing genes to the array; (c) obtaining a biological sample from a subject having an acute respiratory infection (ARI); (d) isolating RNA from the sample to create a transcriptome; (e) measuring the transcriptome on the array (e.g., by measuring fluorescence or current proportional to the level of gene expression, etc.); (f) normalizing the measurements of the transcriptome relative to the normalizing genes, electronically transferring the normalized measurements to a computer to execute the one or more classifiers, (g) generating a report; and optionally, (h) administering appropriate treatment based on the results.

[0032] In some embodiments, the method further comprises externally validating the ARI classifier against a known data set comprising at least two relevant clinical attributes. In some embodiments, the data set is selected from the group consisting of GSE6269, GSE42026, GSE40396, GSE20346, GSE42834, and combinations thereof.

[0033] Yet another aspect of the application provides all that is disclosed and shown herein.

[0034] Also provided is the use of an ARI classifier as taught herein in a method of treating an acute respiratory infection (ARI) of unknown etiology in a subject. BRIEF DESCRIPTION OF DRAWINGS

[0035] The foregoing aspects and other features of the present teachings are explained in the following description, taken in connection with the accompanying drawings, wherein:

[0036] Figure 1is a schematic diagram showing the method (training 10) to obtain classifiers according to some embodiments of the application, where each classifier is composed of a weighted sum of all or a subset of normalized gene expression levels. This weighted sum defines the probability that allows a decision (classification), especially when compared to a threshold or confidence interval. The exact gene combination of each classifier obtained by training, their weights and threshold are specific to the particular platform. The classifiers (or more precisely, their elements, i.e. weights and threshold or confidence interval (value)) go into a database. The weights with non-zero values determine the subset of genes used by the classifier. Repeated within the designated platform matching the gene expression values to obtain all three classifiers (bacterial ARI, viral ARI and non-infectious ARI).

[0037] Figure 2 is a diagram showing an example of generating and / or using classifiers according to some embodiments of the application.

[0038] Figure 3 is a schematic diagram showing the method of classification 20 of the etiology of an acute respiratory symptom suffered by a subject using classifiers according to some embodiments of the application.

[0039] Figure 4 is provided a schematic diagram showing the decision pattern of determining the etiology of an ARI in a subject using secondary classifiers according to some embodiments of the application.

[0040] Figure 5 is a diagram of the example training method given in Example 1. A classifier for each case was developed using patient cohorts comprising bacterial ARI, viral ARI, or non-infectious disease. Leave one out cross-validation was used to validate the combined ARI classifier and to compare it to three published classifiers of bacterial versus viral infection. The combined ARI classifier was also externally validated in six publicly available datasets. In one experiment, healthy volunteers were included as "non-infectious" controls in the training set set to determine their suitability. All subsequent experiments were performed without using this cohort of healthy subjects.

[0041] Figure 6 is provided a diagram showing the results of leave one out cross-validation of the three classifiers (bacterial ARI, viral ARI and non-infectious disease) according to the example training method provided in Example 1. Each patient was assigned a probability of having a bacterial ARI (triangles), a viral ARI (circles), and a non-infectious disease (squares). Patients clinically determined to have a bacterial ARI, a viral ARI, or a non-infectious disease are given in the upper, middle and lower panels, respectively. The overall classification accuracy was 87%.

[0042] Figure 7 is a plot showing the evaluation of healthy adults as non-infectious controls compared to sick but non-infectious controls. This plot demonstrates the unexpected advantage of using sick but non-infectious subjects as controls.

[0043] Figure 8 shows the positive and negative predictive values for A) bacterial and B) viral ARI classification as a function of prevalence.

[0044] Figure 9 is a Venn diagram representing the overlap in bacterial ARI, viral ARI, and non-infectious disease classifiers. There are 71 genes in the bacterial ARI classifier, 33 genes in the viral ARI classifier, and 26 genes in the non-infectious disease classifier. There is one gene overlap between the bacterial and viral ARI classifiers. There are five genes overlap between the bacterial ARI and non-infectious disease classifiers. There are four genes overlap between the viral ARI and non-infectious disease classifiers.

[0045] Figure 10 is a plot showing the classifier performance in patients with co-infections with the aid of bacterial and viral pathogen identification. The bacterial and viral ARI classifiers were trained on subjects with either bacterial (n=22) or viral (n=71) infections (GSE60244). This same dataset also included 25 subjects with bacterial / viral co-infections. As shown in the plot, the bacterial and viral classifier predictions were normalized to the same scale. Each subject received two probabilities: the probability of a bacterial ARI host response and the probability of a viral ARI host response. A probability score of 0.5 or above was considered positive. Subjects 1-6 had a bacterial host response. Subjects 7-9 had both a bacterial and viral host response that could indicate a true co-infection. Subjects 10-23 had a viral host response. Subjects 24-25 had neither a bacterial or viral host response.

[0046] Figure 11is a block diagram of a classification system and / or computer program product that can be used in a platform. Classification system and / or computer program product 1100 can include a processor subsystem 1140, which includes one or more central processing units (CPUs) on which one or more operating systems and / or one or more application programs run. Although one processor 1140 is shown, it is understood that multiple processors 1140 can be present, which can be electrically connected to one another or separate. One or more processors 1140 are configured to execute computer program code from a memory device, such as memory 1150, to perform at least some of the operations and methods described herein. Storage circuitry 1170 can store a database that provides access to data / parameters / classifiers used by classification system 1110, such as identification signatures, weights, thresholds, etc. Input / output circuitry 1160 can include a display and / or user input devices, such as a keyboard, touch screen, and / or pointing device. Devices connected to input / output circuitry 1160 can be used to provide information to processor 1140 by a user of classification system 1100. Devices connected to input / output circuitry 1160 can include a network or communication controller, input devices (keyboard, mouse, touch screen, etc.) and output devices (printer or display). An optional upgrade circuitry 1180 can be included as an interface for providing upgrades to classification system 1100, such as upgrades to code executed by processor 1140 stored in memory 1150 and / or storage circuitry 1170. Upgrades provided through upgrade circuitry 1180 can also include upgrades to portions of storage circuitry 1170 related to databases and / or other data storage forms that maintain information for classification system 1100, such as identification signatures, weights, thresholds, etc. Sample input circuitry 1110 provides an interface for classification system 1100 to receive biological samples to be analyzed. Sample processing circuitry 1120 can also process biological samples within classification system 1100 to prepare biological samples for automated analysis. DETAILED DESCRIPTION

[0047] For the purposes of promoting an understanding of the principles of the application, reference will now be made to the preferred embodiments and will be described in specific language. It will be understood that such alterations and further modifications of the application illustrated and further improvements thereto are contemplated as falling within the scope of the application as defined throughout this disclosure.

[0048] The articles "a" and "an" as used herein mean one or more than one (i.e., at least one) of the grammatical object of the article. By way of example, "an element" means one element or more than one element.

[0049] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0050] The present application provides that changes in gene, protein, and metabolite expression in blood in response to pathogenic exposure leading to acute respiratory infection can be used to identify and characterize the etiology of ARI in a subject with a high degree of accuracy.

[0051] Definitions

[0052] As used herein, the term "acute respiratory infection" or "ARI" refers to an upper or lower respiratory tract infection or disease, usually attributed to a bacterial or viral pathogen, that exhibits symptoms and / or physical findings consistent with infection (e.g., symptoms such as cough, wheezing, fever, sore throat, congestion; physical findings such as elevated heart rate, elevated respiratory rate, abnormal white blood cell count, low arterial carbon dioxide tension (PaC02), etc.) and is characterized by rapid symptom development over hours to days. An ARI can be primarily of the upper respiratory tract (URI), lower respiratory tract (LRI), or a combination of both. An ARI can have systemic effects due to spread of the infection through the respiratory tract or due to collateral damage induced through the immune response. Examples of the former include Staphylococcus aureus pneumonia, which spreads to the bloodstream and can cause secondary sites of infection, including endocarditis (infection of heart valves), septic arthritis (infection of joints), or osteomyelitis (infection of bone). Examples of the latter include influenza pneumonia leading to acute respiratory distress syndrome and respiratory failure.

[0053] The term "signature" as used herein refers to a set of biological analytes and measurable amounts of the analytes whose specific combination indicates the presence or absence of a specified biological state. These signatures are found in a plurality of subjects having a known state (e.g., having a confirmed respiratory bacterial infection, respiratory viral infection, or suffering from a non-infectious disease) and discriminate (alone or in combination) for one or more target categories or outcomes. These measurable analytes, also known as biomarkers, can be, but are not limited to, gene expression levels, protein or peptide levels, or metabolite levels. See also Courchesne et al. US 2015 / 0227681; Eden et al. US 2016 / 0153993.

[0054] In some embodiments as disclosed herein, the "signature" is a specific combination of genes whose expression levels, when combined into a classifier as taught herein, distinguish a condition such as bacterial ARI, viral ARI, or non-infectious disease. See, e.g., Tables 1, 2, 9, 10, and 12, infra. In some embodiments, the signature is agnostic to the species of respiratory virus or bacteria (i.e., does not distinguish between specific genera or species of virus or bacteria in differentiating between virus or bacteria), and / or is agnostic to the specific cause of non-infectious disease.

[0055] As used herein, the terms "classifier" and "predictor" are used interchangeably and refer to a mathematical function that uses the values of the signature (e.g., gene expression levels for a defined set of genes) and predetermined coefficients (or weights) for each signature element to generate a score for a given observation or individual patient for the purpose of assignment to a category. The classifier can be linear and / or probabilistic. A classifier is linear if the score is a function of a weighted sum of the signature values by a set of coefficients. In addition, a classifier is probabilistic if the function of the signature values generates a probability, a value between 0 and 1.0 (or 0 and 100%), respectively quantifying the likelihood that the subject or observation belongs to a particular category or will have a particular outcome. Probit regression and logistic regression are examples of probabilistic linear classifiers that generate probabilities using the probit and logistic link functions, respectively.

[0056] A classifier as taught herein can be obtained by a process known as "training" that utilizes a set of data containing observations with known category membership (e.g., bacterial ARI, viral ARI, and / or non-infectious disease). See Figure 1 Specifically, training seeks to find the best coefficients (i.e., weights) for each element of a given signature (e.g., gene expression level elements), as well as the best signature, where the best result is determined by the highest available classification accuracy.

[0057] "Classification" refers to the process of assigning one or more categories or outcomes to a subject with acute respiratory symptoms or at risk thereof (e.g., a patient is infected or not infected by a pathogen, another categorization can be a patient is infected by a virus and / or by a bacteria). See Figure 3 In some cases, a subject can be classified into more than one category, for example in the case of a combined bacterial and viral infection. The outcome or category is determined by the score value provided by the classifier, which can be compared to a cutoff or threshold value, a confidence level, or a limit. In other cases, a probability of belonging to a particular category can be given (e.g., if the classifier reports a probability).

[0058] As used herein, the term "indicative" when used in conjunction with gene expression levels means up-regulated or down-regulated, altered, or modified gene expression levels as compared to expression levels in an alternative biological state (e.g., bacterial ARI or viral ARI) or a control. When used in conjunction with protein levels, the term "indicative" means higher or lower, increased or decreased, altered, or modified protein levels as compared to one or more standard protein levels in an alternative biological state.

[0059] The terms "subject" and "patient" are used interchangeably and refer to any animal which is the object of examination, study or treatment. It is not intended that the present application be limited to any particular type of subject. In some embodiments of the present application, humans are the preferred subjects, while in other embodiments non-human animals are the preferred subjects, including, but not limited to, mice, monkeys, ferrets, cows, sheep, goats, pigs, chickens, turkeys, dogs, cats, horses, and reptiles. In certain embodiments, the subject has an ARI or displays symptoms similar to an ARI.

[0060] "Platform" or "technology" as used herein refers to the apparatus (e.g., instruments and associated components, computers, computer readable media including one or more databases as taught herein, reagents, etc.) that can be used to measure the identifying markers, e.g., gene expression levels, according to the present application. Examples of platforms include, but are not limited to, array platforms, thermal cycler platforms (e.g., multiplex and / or real-time PCR platforms), nucleic acid sequencing platforms, hybridization and multi-signal encoding (e.g., fluorescent) detector platforms, etc., nucleic acid mass spectrometry platforms, magnetic resonance platforms, and combinations thereof.

[0061] In some embodiments, the platform is configured to semi-quantitatively measure gene expression levels, i.e., measure expression levels as estimates and / or relative to one another or to one or more specified markers (e.g., expression of another "standard" or "reference" gene) as compared to measuring in discrete or absolute expression.

[0062] In some embodiments, semi-quantitative measurement includes "real-time PCR" by performing PCR cycles until a signal indicative of the specified mRNA is detected and using the number of PCR cycles required until detection is made to provide an estimated or relative expression level of the gene within the identifying marker.

[0063] Real-time PCR platforms include, for example, TaqMan® Low Density Arrays (TLDA), in which samples undergo multiplexed reverse transcription followed by real-time PCR on an array card with a set of wells in which real-time PCR is performed. See Kodani et al., 2011, J. Clin. Microbiol. 49(6):2175-2182. Real-time PCR platforms also include, for example, Biocartis Idylla TM Sample-to-result technologies, in which cells are lysed, DNA / RNA is extracted, and real-time PCR is performed and results are detected.

[0064] Magnetic resonance platforms include, for example, T2 Biosystems® T2 Magnetic Resonance (T2MR®) technology, in which molecular targets can be determined in a biological sample without the need for purification.

[0065] The terms "array," "microarray," and "microarray" are interchangeable and refer to an arrangement of a set of nucleotide sequences given on a substrate. Any type of array can be used in the methods provided herein. For example, the array can be on a solid substrate such as a glass slide (solid phase array), or on a semi-solid substrate such as a nitrocellulose membrane. The array can also be given on a bead, i.e., a bead array. The beads are typically microscopic and can be made of, for example, polystyrene. The array can also be given on a nanoparticle, which can be made of, for example, gold, among others, but also silver, palladium, or platinum. See, for example, the Nanosphere Verigene® system, which uses gold nanoparticle probe technology. Magnetic nanoparticles can also be used. Other examples include nuclear magnetic resonance microcoils. The nucleotide sequences can be DNA, RNA, or any arrangement thereof (e.g., nucleotide analogs such as locked nucleic acids (LNAs), etc.). In some embodiments, the nucleotide sequences span exon / intron boundaries to detect spliced or mature RNA species rather than genomic DNA gene expression. The nucleotide sequences can also be partial sequences from genes, primers, whole gene sequences, non-coding sequences, coding sequences, published sequences, known sequences, or new sequences. The array can additionally include other compounds that specifically bind proteins or metabolites, such as antibodies, peptides, proteins, tissues, cells, chemicals, carbohydrates, etc.

[0066] Array platforms include, for example, the above-mentioned TaqMan® Low Density Arrays (TLDA), as well as Affymetrix® microarray platforms.

[0067] Hybridization and multiple signal encoding detector platforms include, for example, NanoString nCounter® technology, in which hybridization of color-coded barcodes attached to target-specific probes (e.g., corresponding to a gene expression transcript) is detected; and Luminex® xMAP® technology, in which microsphere beads are color-coded and coated with target-specific (e.g., gene expression transcript) probes for detection; and Illumina® BeadArray, in which microbeads are assembled onto a fiber-optic bundle or a planar silica glass slide and coated with target-specific (e.g., gene expression transcript) probes for detection.

[0068] Nucleic acid mass spectrometry platforms include, for example, Ibis Biosciences Plex-ID® detector, in which amplified DNA is detected using DNA mass spectrometry with mass profiles.

[0069] Thermal cycler platforms include, for example, FilmArray® multiplex PCR system, which extracts and purifies nucleic acids from an unprocessed sample and performs nested multiplex PCR; and RainDrop digital PCR system, which is a droplet-based PCR platform using a microfluidic chip.

[0070] The term "computer readable medium" refers to any device or system for storing and providing information (e.g., data and instructions) to a computer processor. Examples of computer readable media include, but are not limited to, DVDs, CDs, hard drives, magnetic tape, and servers for streaming media over a network, as well as applications such as those found on smartphones and tablets. In various embodiments, various aspects of the present application, including data structures and methods, can be stored on a computer readable medium. Processing and data can also be carried out on many device types, including but not limited to desktop and notebook computers, tablets, smartphones, and the like.

[0071] As used herein, the term "biological sample" includes any sample that can be taken from a subject containing genetic material that can be used in the methods provided herein. For example, a biological sample can include a peripheral blood sample. The term "peripheral blood sample" refers to a blood sample taken from the systemic system of the body that is circulating in the circulatory system or body. Other samples can include those taken from the upper respiratory tract, including but not limited to sputum, nasopharyngeal swab, and nasopharyngeal wash. Biological samples can also include those taken from the lower respiratory tract, including but not limited to bronchoalveolar lavage fluid and endotracheal aspirate. Biological samples can also include any combination thereof.

[0072] The term "genetic material" refers to the material used to store genetic information in the nucleus or mitochondria of a cell of an organism. Examples of genetic material include, but are not limited to, double-stranded and single-stranded DNA, cDNA, RNA, and mRNA.

[0073] The term "a plurality of nucleic acid oligomers" refers to two or more nucleic acid oligomers, which can be DNA or RNA.

[0074] As used herein, the terms "treat," "treatment," and "treating" refer to the reduction or amelioration of the severity, duration, and / or progression of a disease or disorder or one or more symptoms thereof resulting from the administration of one or more therapies. Such terms refer to a reduction in replication of a virus or bacteria, a reduction in transmission of a virus or bacteria to other organs or tissues in a subject or to other subjects. Treatment can also include therapies for ARIs caused by non-infectious diseases, such as allergy treatments, asthma treatments, and the like.

[0075] The term "effective amount" refers to the amount of a therapeutic agent sufficient to exert a physiological effect in a subject. The term "responsivity" refers to the change in the level of gene expression of a gene in a subject in response to the subject being infected with a virus or bacteria or suffering from a non-infectious disease as compared to the level of gene expression of the gene in a subject or control subject that is not infected with a virus, bacteria, or suffers from a non-infectious disease.

[0076] The term "appropriate treatment regimen" refers to the standard of care required to treat a particular disease or condition. Typically, such a regimen entails the act of administering to the subject one or more therapeutic agents capable of producing a curative effect in the disease state. For example, a therapeutic agent for treating a subject with bacteremia is an antibiotic, including but not limited to penicillin, cephalosporin, fluroquinolone, tetracycline, macrolide, and aminoglycoside. Therapeutic agents for treating a subject with a viral respiratory infection include but are not limited to oseltamivir, an RNAi antiviral, inhaled ribavirin, monoclonal antibody respigam, zanamivir, and a neuraminidase blocker. The present invention contemplates the use of the methods of the present invention to identify treatments with yet unavailable antivirals or antibiotics. Appropriate treatment regimens also include treatments for ARIs caused by non-infectious diseases, such as the administration of allergy treatments including but not limited to antihistamines, decongestants, anticholinergic nasal sprays, leukotriene inhibitors, mast cell inhibitors, steroid nasal sprays, and the like; and asthma treatments including but not limited to inhaled corticosteroids, leukotriene modulators, long-acting beta agonists, combinations of inhaled agents (e.g., fluticasone-salmeterol; budesonide-formoterol; mometasone-formoterol; etc.), theophylline, short-acting beta agonists, ipratropium, oral and intravenous corticosteroids, omalizumab, and the like.

[0077] Typically, such a regimen entails the act of administering to the subject one or more therapeutic agents capable of producing a reduction in symptoms associated with the disease state. Examples of such therapeutic agents include but are not limited to NSAIDS, acetaminophen, antihistamines, beta-agonists, anti-tussives, or other drugs that reduce symptoms associated with the disease process.

[0078] Method of generating a classifier (training)

[0079] The present invention provides methods of generating classifiers (also referred to as training 10) for methods of determining the etiology of acute respiratory disease in a subject. Classifiers based on gene expression are developed that can be used to identify and characterize the etiology of ARI in a subject with a high degree of accuracy.

[0080] Thus, and as shown in Figure 1 , one aspect of the present invention provides a method of manufacturing an acute respiratory infection (ARI) classifier, the method comprising, consisting of, or consisting essentially of: (i) obtaining biological samples (e.g., peripheral blood samples) from a plurality of subjects having a bacterial, viral, or non-infectious acute respiratory infection 100; (ii) optionally, isolating RNA from the samples (e.g., total RNA forming a transcriptome) 105 (not shown in Figure 1 ); (iii) measuring gene expression levels of a plurality of genes 110 (i.e., some or all of the genes expressed in the RNA); (iv) normalizing the gene expression levels 120; and (v) generating a bacterial ARI classifier, a viral ARI classifier, or a non-infectious disease classifier based on the results 130.

[0081] In some embodiments, the sample is not purified after collection. In some embodiments, the sample can be purified prior to or after cell lysis to remove extraneous material. In some embodiments, the sample is purified with cell lysis and removal of cellular material, isolation of nucleic acids, and / or reduction of bulk transcripts such as globin or ribosomal RNA.

[0082] In some embodiments, measuring gene expression levels can include generating one or more microarrays using the transcriptome; measuring the transcriptome using a plurality of primers; analyzing and correcting for batch differences.

[0083] In some embodiments, the method further comprises uploading 140 a final gene target list regarding the generated classifier, associated weights (w n ), and thresholds to one or more databases.

[0084] Examples of generating the classifiers are detailed in Figure 2 . As shown in Figure 2As shown in the middle, a gene expression-based classifier is formed for each case (i.e., bacterial acute respiratory infection, viral acute respiratory infection, or non-infectious etiology) using biological samples from patient cohorts comprising bacterial ARI, viral ARI, or non-infectious disease. Specifically, a bacterial ARI classifier is obtained to positively identify those patients with bacterial ARI over viral ARI or non-infectious disease. A viral ARI classifier is obtained to positively identify those patients with viral ARI over bacterial ARI or non-infectious disease (NI). A non-infectious disease classifier is generated to improve the specificity of the bacterial and viral ARI classifiers. Next, signatures are generated for the bacterial ARI classifier, viral ARI classifier, and non-infectious disease classifier (e.g., by applying a sparse logistic regression model).

[0085] If desired, the three classifiers can then be combined into a single classifier, referred to as the "ARI classifier," by a one-versus-all scheme in which the class label is assigned by the largest member probability. See also Figure 5 In some embodiments, the combined ARI classifier can be validated in the same population from which it came using leave-one-out cross-validation, and / or in some embodiments, can be validated using publicly available human gene expression datasets of samples from subjects with diseases of known etiology. For example, validation can be performed using publicly available human gene expression datasets (e.g., GSE6269, GSE42026, GSE40396, GSE20346, and / or GSE42834) selected insofar as they comprise at least two clinical groups (bacterial ARI, viral ARI, or non-infectious disease).

[0086] The classifiers can be validated in a set of standard samples from subjects with diseases of known etiology (i.e., bacterial ARI, viral ARI, or non-infectious disease).

[0087] The methods described herein for training can be readily converted by one of ordinary skill in the art to different gene expression detection (e.g., mRNA detection and quantification) platforms.

[0088] The methods and assays of the present application can be based on gene expression, for example, by direct measurement of RNA, measurement of derivative material (e.g., cDNA), and measurement of RNA products (e.g., encoded proteins or peptides). Any method of extracting and screening gene expression can be used and is within the scope of the present application.

[0089] In some embodiments, the measurement comprises detection and quantification (e.g., semi-quantification) of mRNA in the sample. In some embodiments, the gene expression levels are adjusted relative to the level of one or more standard genes ("normalized"). As known in the art, normalization is done to remove technical variability inherent to the platform, resulting in an amount or relative amount (e.g., of expressed genes).

[0090] In some embodiments, detection and quantification of mRNA can first comprise a reverse transcription and / or amplification step, such as RT-PCR, e.g., quantitative RT-PCR. In some embodiments, the detection and quantification can be based on unamplified mRNA molecules present in or purified from the biological sample. Direct detection and measurement of RNA molecules typically involves hybridization to complementary primers and / or labeled probes. Such methods include conventional Northern blotting and surface-enhanced Raman spectroscopy (SERS), which involves illuminating a sample exposed to a surface of a plasmon-active metal structure with a laser and measuring changes in the light frequency as it is scattered.

[0091] Similarly, detection of RNA derivatives such as cDNA typically involves hybridization to complementary primers and / or labeled probes. This can include high-density oligonucleotide probe arrays (e.g., solid-state microarrays and bead arrays) or related probe-hybridization methods, as well as polymerase chain reaction (PCR)-based amplification and detection for relative and absolute quantification of specific RNA molecules, including real-time, digital, and end-point PCR methods.

[0092] In addition, sequencing-based methods can be used to detect and quantify levels of RNA or RNA-derived material. When applied to RNA, sequencing methods are referred to as RNAseq, and provide both qualitative (sequence in the sample, or presence / absence of RNA, or its cognate cDNA) and quantitative (copy number) information about RNA molecules from a sample. See, e.g., Wang et al., 2009 Nat. Rev. Genet. 10(1):57-63. Another sequence-based method, serial analysis of gene expression (SAGE), uses cDNA "tags" as a proxy for measuring expression levels of RNA molecules.

[0093] In addition, specialized platforms for mRNA detection and quantification can also be used to accomplish the methods of the present application. Examples of these are the Pixel TM TM ​Systems, NanoString® Technologies nCounter Gene Expression System; mRNA-Seq, Tag-Profiling, BeadArray™ technology from Illumina and VeraCode, ICEPlex system from PrimeraDx, and QuantiGene 2.0 multiplex assay from Affymetrix.

[0094] For example, RNA preservation reagents such as PAXgene TM RNA tubes (PreAnalytiX, Valencia, Calif.) can be used to collect RNA from whole blood from a subject. Standard PAXgene TM or Versagene TM RNA extraction protocols can be used to extract RNA. Versagene TM kits produce higher yields of higher quality RNA from PAXgene TM RNA tubes. After RNA extraction, GLOBINCIear TM (Ambion, Austin, Tex.) can be used for globin protein reduction. (This method uses a bead-oligonucleotide construct to bind globin mRNA, and in our experience we are able to remove over 90% of globin mRNA.) Removal of large amounts of transcripts that are not of interest can increase sensitivity of assays such as microarray platforms, depending on the technology.

[0095] RNA quality can be assessed by a number of means. For example, RNA quality can be assessed immediately after extraction using an Agilent 2100 Bioanalyzer. This analysis provides a RNA Integrity Number (RIN) as a quantitative measure of RNA quality. In addition, after globin reduction, samples can be compared to globin reduction standards. In addition, scaling factors and background can be assessed after hybridization to a microarray.

[0096] Gene expression from whole blood samples can be rapidly identified using real-time PCR. For example, isolated RNA can be reverse transcribed and then amplified and detected in real-time using either a non-specific fluorescent dye intercalated into the resulting ds-DNA or sequence-specific DNA probes labeled with a fluorescent reporter protein that only allows detection after the probe is hybridized to its complementary DNA target.

[0097] Thus, it is understood that there are many methods of mRNA quantification and detection that can be used by the platform according to the methods disclosed herein.

[0098] Typically following detection and quantification, expression levels are appropriately normalized using methods typically practiced by those of ordinary skill in the art for the particular platform.

[0099] In cases where mRNA detection and quantification and matching normalization methods are appropriate for the platform, it is merely a matter of using patient samples carefully selected and determined for the training method. For example, using cohort generation described below, appropriate weighting values (coefficients) are generated to be used in conjunction with the genes in the three signature genes in the classifier for the platform. These subject-sample generated coefficients and cutoffs can also be used for trials performed using different mRNA detection and quantification platforms.

[0100] In some embodiments, the individual categories of the classifier (i.e., bacterial ARI, viral ARI, non-infectious disease) are formed from cohorts comprising multiple such causes thereof. For example, the bacterial ARI classifier is obtained from a cohort with bacterial infections from multiple bacterial genera and / or species, the viral ARI classifier is obtained from a cohort with viral infections from multiple viral genera and / or species, and the non-infectious disease classifier is obtained from a cohort with non-infectious diseases due to multiple non-infectious causes. See, e.g., Table 8. In this way, the respective classifiers obtained are agnostic to the underlying bacterial, viral, and non-infectious causes. In some embodiments, some or all of the subjects in the cohort with non-infectious etiology have symptoms consistent with a respiratory infection.

[0101] In some embodiments, the signature genes can be obtained using a supervised statistical method known as sparse linear classification, in which sets of genes are identified by the model according to their ability to separate phenotypes during training using selected sets of patient samples. The result of the training is the gene signature and classification coefficients for the three comparisons. Together, the signature genes and coefficients provide the classifier or predictor. Training can also be used to establish threshold or cutoff values. Threshold or cutoff values can be adjusted to change trial performance, e.g., trial sensitivity and specificity. For example, the threshold for bacterial ARI can be intentionally lowered to increase the sensitivity of the trial for bacterial infections, if desired.

[0102] In some embodiments, the classifier generating iteratively comprises: (i) assigning a weight to each normalized gene expression value, inputting the weight and expression value for each gene into a classifier (e.g., a linear regression classifier) equation and determining a score of the outcome for each of the plurality of subjects, then (ii) determining the accuracy of the classification of each outcome among the plurality of subjects, and then (iii) adjusting the weights until the accuracy of the classification is optimized. Genes with non-zero weights are included in the respective classifier.

[0103] In some embodiments, the classifier is a linear regression classifier and the generating comprises converting the score of the classifier into a probability using a link function. As known in the art, a link function specifies the relationship between the target / output of a model (e.g., the probability of a bacterial infection) and the systematic elements of a linear model (in this case, the combination of explanatory variables including predictors). It states how the expected value of the response relates to the linear predictors of the explanatory variables.

[0104] Method of classification

[0105] The present invention also provides methods for determining whether a patient has a respiratory disease due to a bacterial infection, a viral infection, or a non-infectious cause. The methods for making such a determination rely on the use of a classifier obtained as taught herein. The methods can comprise: a) measuring the expression level of a predefined plurality of sets of genes (i.e., for one or more of the three identification signatures); b) normalizing the gene expression levels to the technique used to make the measurements; c) taking those values and inputting them into a bacterial, viral, and / or non-infectious disease classifier (i.e., predictor) having a predefined weighting value (coefficient) for each of the genes in each identification signature; d) comparing the output of the classifier to a predefined threshold, cutoff, confidence interval, or range of values indicating the likelihood of infection; and optionally e) reporting the results of the classifiers collectively.

[0106] A brief summary of such a method is provided in Figure 3 In this illustration, each of the three gene identification signatures provides information about the host response to a different ARI etiology (bacterial or viral) or to a sick but not infectious state (NI) in the patient. These identification signatures are sets of gene transcripts that have increased or decreased expression levels in response to one of three clinical states: bacterial ARI, viral ARI, or non-infectious but sick state. These identification signatures are obtained (trained 10) using careful adjudication of patient samples with one or more of the target conditions.

[0107] Reference is made to Figure 3After a biological sample (e.g., a blood sample) is obtained from a patient, mRNA is extracted in some embodiments. The mRNA (or defined regions of each mRNA) is quantified for all or a subset of the genes in the signature. Depending on the device used for quantification, it can first be necessary to purify the mRNA from the sample.

[0108] The signatures identify clinical states and are defined with respect to at least one of the other two possibilities. For example, the bacterial ARI signature is identified as a set of biomarkers (here, represented by mRNA transcripts of genes) that distinguish patients with bacterial ARI from patients without bacterial ARI (including patients with viral ARI or non-infectious disease involving the application). The viral ARI signature is defined by a set of biomarkers that distinguish patients with viral ARI from patients without viral ARI (including patients with bacterial ARI or non-infectious disease). The non-infectious disease signature is defined by a set of biomarkers that distinguish patients with non-infectious etiology from patients with bacterial or viral ARI.

[0109] The normalized expression level of each gene of the signature (e.g., first column of Table 9) is an explanatory or independent variable or feature used in the classifier. For example, the classifier can have the general form of a probit regression equation:

[0110]

[0111] where the condition is bacterial ARI, viral ARI, or non-infectious disease; Φ(.) is the probit (or logistic) link function; {β1, β2,..., β d} are coefficients obtained during training (e.g., second, third, and fourth columns from Table 9) (coefficients can also be denoted herein as {w1, w2,..., w d} as "weights"); {X1, X2,..., X d} are the normalized gene expression levels of the signature; and d is the size of the signature (i.e., the number of genes).

[0112] As will be appreciated by those skilled in the art, the value of the coefficient for each explanatory variable will vary for each technical platform used to measure the expression of the genes or subset of genes used in the probit regression model. For example, the coefficient for each of the features in the classifier algorithm for gene expression measured by Affymetrix U133A 2.0 microarrays is shown in Table 9.

[0113] The sensitivity, specificity, and overall accuracy of each classifier can be optimized by changing the threshold of classification using the receiving operating characteristic (ROC) curve.

[0114] Another aspect of the application provides a method for determining whether an acute respiratory infection (ARI) in a subject is of bacterial origin, viral origin, or non-infectious origin, the method comprising, consisting of, or consisting essentially of: a) obtaining a biological sample from the subject; b) determining the gene expression profile of the subject by evaluating the expression levels of a predefined plurality of sets of genes (i.e., three signatures) from the biological sample; c) normalizing the gene expression levels to generate normalized values for the technique used to make the measurements; d) inputting the normalized values into a bacterial, viral, and non-infectious disease classifier (i.e., predictor) having a predefined weighting value (coefficient) for each of the genes in each signature; e) comparing the output of the classifier to a predefined threshold, cutoff, or range of values indicative of the likelihood of infection; and e) classifying the sample as a bacterial etiology, viral etiology, or non-infectious disease. In some embodiments, the method further comprises generating a report assigning a score to the patient indicative of the probability of the etiology of the ARI.

[0115] The classifiers developed during the training and training set of samples are applied for predictive purposes to diagnose new individuals ("classification"). For each subject or patient, a biological sample is taken and the normalized expression level (i.e., relative mRNA expression) in the sample for each of the genes specified by the signatures discovered during training is input into the classifier. For each gene, the classifier also uses the weighting coefficient discovered during training. As output, the classifier is used to calculate three probability values. Each probability value can be used to determine the likelihood of three considered clinical states: bacterial ARI, viral ARI, and non-infectious disease.

[0116] In some embodiments, the results of each of the classifiers are reported - the probability that the new subject or patient has a bacterial ARI, a viral ARI, or a non-infectious disease. Finally, the three signatures with their respective coefficients are applied to the individual patient to obtain three probability values, i.e., the probability of having a bacterial ARI, a viral ARI, and a non-infectious disease. In some embodiments, these values can be reported relative to a reference range that represents the confidence with which the classification was made. In some embodiments, the output of the classifiers can be compared to a threshold, e.g., to report "positive" if the classifier score or probability exceeds a threshold that indicates the presence of one or more of a bacterial ARI, a viral ARI, or a non-infectious disease. If the classifier score or probability fails to reach the threshold, the result will be reported as "negative" for the corresponding condition. Optionally, only the values for bacterial and viral ARIs are reported, and the possibility of not mentioning a disease but being non-infectious is reported.

[0117] It should be noted that classifiers obtained with one platform can show the best performance on another platform. This can be due to a mix of platform-specific probe or other technical issues. Therefore, methods are also described herein to adapt signatures as taught herein from one platform to another.

[0118] For example, for the Affymetrix platform, signatures obtained from the Affymetrix platform can be adapted to the TLDA platform by using corresponding TLDA probes for the genes in the signature and / or alternative genes related to those in the signature. Table 1 shows a list of Affymetrix probes and the genes they measure, as well as "replacement genes" that are introduced as replacements for gene probes that can not perform well on the TLDA platform due to technical reasons or as replacements for those Affymetrix probes for which no homologous TLDA probe exists. These replacements can represent highly related genes or can be probes that bind to different positions in the same gene transcript. Additional genes can be included, such as pan-viral gene probes. The weights shown in Table 1 are the weights calculated for the classifiers implemented on the microarray platform. Weights that have not been estimated are represented by "NA" in the table. (Example 4 below provides the complete conversion of these classifiers to the TLDA platform.) Reference probes for TLDA (i.e., normalization genes such as TRAP1, PPIB, GAPDH, and 18S) also have "NA" in the column for weights and Affymetrix probe set ID (these are not part of the classifiers). Additional gene probes that do not necessarily correspond to Affymetrix probe sets also have "NA" in the column for Affymetrix probe set ID.

[0119] Table 1 : The primary gene list for the TLDA platform panel is as follows:

[0120] Column 1 Affymetrix Probe Set ID - This is the probe set (primary probe set) identified in the Affy Discovery Analysis

[0121] Columns 2, 3, 4 Estimated Coefficient (Weight) of each probe's contribution to the 3 classifiers from the Affymetrix weights

[0122] Column 5 Gene Name

[0123]

[0124]

[0125]

[0126]

[0127]

[0128] Further discussion of this example identification of markers for the TLDA platform is provided in the following Examples 3 and 4.

[0129] This method of determining the etiology of an ARI can be combined with other tests. For example, if it is determined that a patient has a viral ARI, a subsequent test can be used to determine if influenza A or B can be directly detected or if a host response indicative of such an infection can be detected. Similarly, a subsequent test of the results for a bacterial ARI can be used to determine if a gram-positive or gram-negative bacteria can be directly detected or if a host response indicative of such an infection can be detected. In some embodiments, a simultaneous test can be performed to determine the class of infection using classifiers and also test for a specific pathogen using pathogen-specific probes or detection methods. See, e.g., Eley et al. US 2015 / 0284780 (method for detecting active tuberculosis); Tsalik et al. US 2014 / 0323391 (method for classification of bacterial infection).

[0130] Method of determining a secondary classification of ARI in a subject

[0131] The present application also provides methods of classifying a subject using a two-level classification scheme. Thus, another aspect of the present application provides a method for determining whether an acute respiratory infection (ARI) in a subject is of bacterial origin, viral origin, or non-infectious origin, the method comprising, consisting of, or consisting essentially of: (a) obtaining a biological sample from the subject; (b) determining a gene expression profile of the subject from the biological sample by evaluating the expression levels of a predefined plurality of sets of genes (i.e., three signatures) ; (c) normalizing the gene expression levels as needed for the technology used to make the measurements to generate normalized values; (d) inputting the normalized values into a classifier (i.e., predictor) having a predefined weighting value (coefficient) for each of the genes in each signature; (e) comparing the output of the classifier to a predefined threshold, cutoff, or range of values indicative of the likelihood of infection; (f) repeating step (d) using only the viral and non-infectious classifiers if the sample is negative for bacteria; and (g) classifying the sample as viral etiology or non-infectious disease.

[0132] Another aspect of the present application provides a method for determining whether an acute respiratory infection (ARI) in a subject is of bacterial origin, viral origin, or non-infectious origin, the method comprising, consisting of, or consisting essentially of: (a) obtaining a biological sample from the subject; (b) determining a gene expression profile of the subject from the biological sample by evaluating the expression levels of a predefined plurality of sets of genes (i.e., three signatures) ; (c) normalizing the gene expression levels for the technology used to make the measurements to generate normalized values; (d) inputting the normalized values into a classifier (i.e., predictor) having a predefined weighting value (coefficient) for each of the genes in each signature; (e) comparing the output of the classifier to a predefined threshold, cutoff, or range of values indicative of the likelihood of infection; (f) repeating step (d) using only the bacterial and non-infectious classifiers if the sample is negative for viruses; and (g) classifying the sample as bacterial etiology or non-infectious disease.

[0133] Still another aspect of the application provides a method for determining whether an acute respiratory infection (ARI) in a subject is of bacterial origin, viral origin, or non-infectious origin, the method comprising, consisting of, or consisting essentially of: (a) obtaining a biological sample from the subject; (b) determining a gene expression profile of the subject from the biological sample by evaluating the expression levels of a predefined plurality of sets of genes (i.e., three signatures) ; (c) normalizing the gene expression levels to a technology used to make the measurements to generate normalized values; (d) inputting the normalized values into a classifier (i.e., predictor) having a predefined weighting value (coefficient) for each of the genes in each signature; (e) comparing the output of the classifier to a predefined threshold, cutoff, or range of values indicative of the likelihood of infection; (f) repeating step (d) using only the viral and bacterial classifiers if the sample is negative for non-infectious disease; and (g) classifying the sample as viral or bacterial etiology.

[0134] In some embodiments, the method further comprises generating a report assigning a score to the patient indicative of the probability of the etiology of the ARI.

[0135] In Figure 4 Classification of a patient's status using a two-stage classification scheme is illustrated in FIG. 1. In this example, the bacterial ARI classifier will distinguish between patients with bacterial ARI and those without bacterial ARI, but it can be either viral ARI or non-infectious etiology. Those patients with non-bacterial ARI can then be classified in a second stage to further distinguish between viral ARI and non-infectious disease. This same process of first and second stage classification can also be applied to the viral ARI classifier in the case where a patient is determined not to have a viral infection, but will then be classified a second time as having either bacterial ARI or non-infectious etiology. Likewise, application of the non-infectious disease classifier as a first test will determine whether a patient has such a non-infectious disease, or conversely, a symptom cause that is infectious. A second classification step will determine whether the infection is due to a bacterial or viral pathogen.

[0136] Results from the three first stage and three second stage classifications can be summarized by a person of skill in the art by a variety of techniques (e.g., summing, counting, or averaging) to produce an actionable report for the provider. In some embodiments, the genes used for such a second stage of classification can be some or all of those given in Table 2.

[0137] In such an example, level 1 classification is performed using the three classifiers described above (bacterial classifier, viral classifier, and non-infectious disease classifier). Thereafter, for those patients with non-bacterial infections, a secondary classifier is defined to distinguish between viral ARI and those with non-infectious disease ( Figure 4 Similarly, for those patients with non-viral infections, a new classifier is used to distinguish between viral infection and non-infectious disease ( Figure 4 , middle column), and for those patients not classified as having non-infectious disease in the first step, a new classifier is used to distinguish between viral and bacterial ARI ( Figure 4 , right column).

[0138] In this secondary approach, nine probabilities can be generated, and those probabilities can be combined in a number of ways. Two strategies are described here as ways to reconcile the three sets of predictions, each with probabilities of bacterial ARI, viral ARI, and non-infectious disease. For example: Highest prediction average probability: the probabilities of all predictions for bacterial ARI are averaged, as are the probabilities of all predictions for viral ARI, and similarly, the probabilities of all predictions for non-infectious disease. The maximum average probability represents the diagnosis.

[0139] Maximum number of predictions: Instead of averaging the probabilities of the predictions for each condition, the number of times a particular diagnosis (i.e., bacterial ARI, viral ARI, or non-infectious disease) is predicted for the patient sample is counted. The best case is when the three classification schemes provide the same answer (e.g., bacterial ARI for scheme 1, bacterial ARI for scheme 2, and bacterial ARI for scheme 3). The worst case is when each scheme gives a different diagnosis, resulting in a 3-way tie.

[0140] Using the training set of patient samples described previously, the results of level 1 classification can be, for example, similar to those given in Table 3 (clinical classification given in rows; diagnostic test predictions given in columns).

[0141]

[0142] For example, classification can be implemented as described above and / or as summarized in Table 2. Table 2 summarizes the gene members in three different classification strategies that address the different diagnostic questions. There are a total of 270 probes collectively comprising the three composite classifiers. The first is called BVS (bacterial ARI, viral ARI, SIRS), which is the same as given below in Example 1. These probes are the same as those given in Table 9, which provides the probe / gene weights used for classification. They also correspond to the genes given in Table 10.

[0143] The second is called 2L, for 2 layers or 2 tiers. This is the tiered approach presented in Figure 4

[0144] The third is a one-level classification approach, BVSH, which is similar to BVS, but also includes a population of healthy controls (similarly described in Example 1). This group has been shown to be a poor control for non-infectious, but there are use cases where distinguishing health can be clinically important. For example, this can include a series of measurements that identify a signature and are relevant to recovery. It can also be used to distinguish patients who have been exposed to an infectious agent and are asymptomatic prior to the onset of symptoms. In the BVSH approach, four groups are presented in the training population - those with bacterial ARI, viral ARI, SIRS (non-infectious disease), and healthy. Using these four groups, four different signatures are generated that distinguish each class from all other possibilities.

[0145] Table 2 Legend:

[0146] Probe = Affymetrix Probe ID

[0147] BVS = Three-classifier model trained on patients with bacterial ARI, viral ARI, and non-infectious disease (with respiratory symptoms). 1 indicates that this probe is included in this three-classifier model. 0 indicates that the probe is not present in this classification scheme.

[0148] BVS-BO = Gene or probe included in the bacterial ARI classifier as part of the BVS classification scheme. This classifier specifically distinguishes patients with bacterial ARI from other etiologies (viral ARI or non-infectious disease).

[0149] BVS-VO = Identifies genes included in the viral ARI classifier for BVS-BO except for this column. This classifier specifically distinguishes patients with viral ARI from other etiologies (bacterial ARI or non-infectious disease).

[0150] BVS-SO = Identifies genes included in the non-infectious disease classifier for BVS-BO or BVS-VO except for this column. This classifier specifically distinguishes patients with non-infectious disease from other etiologies (bacterial or viral ARI).

[0151] ​2L refers to a two-level hierarchical classification scheme. In this column, 1 indicates that the specified probe or gene is included in the classification task. This 2-level classification scheme itself consists of three separate hierarchical tasks. The first level applies a one vs. others, where one can be bacterial ARI, viral ARI, or non-infectious disease. If a given subject falls in the "others" category, then a 2-level classification is performed, which distinguishes the remaining possibilities. 2L-SO is directed to the model's 1st level, which determines whether a given subject has a non-infectious disease, followed by SL-BV, which discriminates between bacterial and viral ARI as possibilities. A 1 in these columns indicates that the gene or probe is included in the specified classification model. 2L-BO and 2L-VS perform another 2-level classification scheme. 2L-VO and 2L-SB include the 3rd model in the 2-level classification scheme.

[0152] Finally, BVSH refers to a horizontal classification scheme, which includes healthy individuals in the training cohort, and thus includes a classifier for the healthy state as compared to bacterial ARI, viral ARI, or non-infectious disease. The dark gray BVSH columns identify any genes or probes included in this classification scheme. This scheme itself includes BVSH-BO, BVSH-VO, BVSH-SO, and BVSH-HO, and their respective probe / gene compositions in these columns are indicated by "1".

[0153] Table 2 provides a summary of the use of members of the gene set for constructing viral, bacterial, and non-infectious disease classifiers according to the desired task. A "1" indicates that the member of the gene is included in the classifier.

[0154]

[0155]

[0156]

[0157]

[0158]

[0159]

[0160]

[0161]

[0162]

[0163]

[0164]

[0165]

[0166]

[0167]

[0168]

[0169]

[0170]

[0171]

[0172] Method of treating a subject having ARI

[0173] Another aspect of the application provides a method of treating an acute respiratory infection (ARI) of unknown etiology in a subject, the method comprising, consisting of, or consisting essentially of: (a) obtaining a biological sample from the subject; (b) determining a gene expression profile of the subject from the biological sample by evaluating the expression level of a predefined plurality of sets of genes (e.g., one, two, or more than three identification markers); (c) normalizing the gene expression level as needed for the technique used to make the measurement to generate a normalized value; (d) inputting the normalized value into a bacterial classifier, a viral classifier, and a non-infectious disease classifier (i.e., predictor) having a predefined weighting value (coefficient) for each of the genes in each of the identification markers; (e) comparing the output of the classifiers to a predefined threshold, cutoff, or range of values indicative of the likelihood of infection; (f) classifying the sample as a bacterial etiology, a viral etiology, or a non-infectious disease; and (g) administering to the subject an appropriate treatment regimen as determined by step (f).

[0174] In some embodiments, step (g) comprises administering an antibacterial treatment when the etiology of the ARI is determined to be bacterial. In other embodiments, step (g) comprises administering an antiviral treatment when the etiology of the ARI is determined to be viral.

[0175] After the cause of the ARI has been determined for a subject, she can undergo treatment, such as anti-viral treatment if the ARI is determined to be viral, and / or she can be quarantined from her family during the course of the infection. Alternatively, if the ARI is determined to be bacterial, a bacterial treatment regimen can be administered (e.g., administration of antibiotics). Those subjects classified as non-infectious diseases can be sent home or observed for further diagnosis and treatment (e.g., allergies, asthma, etc.).

[0176] The personnel processing the peripheral blood sample need not be involved in the comparison, however, it is contemplated that the laboratory can communicate the gene expression levels of the classifier to a medical practitioner for the purpose of identifying the cause of the ARI and for administering appropriate treatment. Further, it is contemplated that after examining the patient, the medical professional will order the reagents to obtain a peripheral blood sample, assay the sample for the classifier, and have the reagents report the cause status of the patient to the medical professional. Once the medical professional has the cause of the ARI, the medical professional can then formulate an appropriate treatment and / or quarantine.

[0177] The methods provided herein can be useful for diagnosing the cause of a disease, thereby properly treating the patient and reducing the inappropriate use of antibiotics. Further, the methods provided herein have a variety of other uses, including but not limited to (1) host-based testing to detect individuals who have been exposed to a pathogen and have an impending but asymptomatic disease (e.g., in the case of natural transmission of a disease through a population, but also in the case of bioterrorism); (2) host-based testing to monitor responses to vaccines or drugs in a clinical trial setting or for immune populations; (3) host-based testing to screen for impending disease prior to deployment (e.g., military deployment or boarding a ship in a civilian setting such as a cruise ship); and (4) host-based testing to screen ARI in livestock (e.g., avian flu and other potential pandemic viruses).

[0178] Another aspect of the application provides a kit for determining the cause of an acute respiratory infection (ARI) in a subject, the kit comprising, consisting of, or consisting essentially of (a) means for extracting a biological sample; (b) means for generating one or more arrays consisting of a plurality of synthetic oligonucleotides having a region homologous to a set of gene transcripts as taught herein; and (c) instructions for use.

[0179] Still another aspect of the application provides a method of using the kit to evaluate the acute respiratory infection (ARI) classifier, the method comprising, consisting of, or consisting essentially of: (a) generating one or more arrays consisting of a plurality of synthetic oligonucleotides having regions homologous to a set of gene transcripts as taught herein; (b) adding oligonucleotides having regions homologous to a normalization gene to the array; (c) obtaining a biological sample from a subject having an acute respiratory infection (ARI); (d) isolating RNA from the sample to create a transcriptome; (e) measuring the transcriptome on the array; (f) normalizing the measurements of the transcriptome relative to the normalization gene, electronically transferring the normalized measurements to a computer to execute the one or more classifier algorithms, (g) generating a report; and optionally, (h) administering appropriate treatment based on the results.

[0180] Classification system

[0181] Referring to Figure 11 , in accordance with the embodiments described herein, the classification system and / or computer program product 1100 can be implemented in or through a platform. The classification system and / or computer program product 1100 can be implemented as one or more enterprise, application, personal, pervasive, and / or embedded computer systems operable to receive, transmit, process, and store data using any suitable combination of software, firmware, and / or hardware and can be standalone and / or interconnected through any conventional, public or proprietary, real and / or virtual, wired and / or wireless network including all or part of the global communication network known as the Internet and can include a plurality of types of tangible, non-transitory computer readable media.

[0182] As shown in Figure 11 , the classification system 1100 can include a processor subsystem 1140 including one or more central processing units (CPUs) on which one or more operating systems and / or one or more application programs are run. Although one processor 1140 is shown, it is understood that multiple processors 1140 can be present, which can be electrically connected to one another or separate. The one or more processors 1140 are configured to execute computer program code from a memory device, such as memory 1150, to perform at least some of the operations and methods described herein and can be any conventional or special purpose processor, including but not limited to digital signal processors (DSPs), field programmable gate arrays (FPGA), application specific integrated circuits (ASICs), and multi-core processors.

[0183] Memory subsystem 1150 can include a memory device layer such as random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM) or flash memory, and / or any other solid state memory device. Storage subsystem 1170, which can include, for example, a portable computer disk drive, a hard disk drive, a portable compact disc read only memory (CD-ROM), DVD, and / or any other type of optical disk storage, magnetic storage device, and / or any other mass storage device known in the art. Storage subsystem 1170 can provide non-volatile storage for data / parameters / classifiers of classification system 1100. Storage subsystem 1170 can include a disk drive and / or a network storage component. Storage subsystem 1170 can be used to store data and / or code for use by processors 1140. In some embodiments, storage subsystem 1170 can store a database that provides access to data / parameters / classifiers such as identification signatures, weights, thresholds, etc. used by classification system 1110. Storage subsystem 1170 can use any combination of one or more computer readable media. Computer readable media can be computer readable signals or computer readable storage media. Computer readable storage media can be, for example but not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage media would include the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. As used in this document, computer readable storage media is any tangible media that can contain or store program instructions for use by or in connection with an instruction execution system, apparatus, or device.

[0184] Input / output subsystem 1160 can include a display and / or user input devices such as a keyboard, touchscreen, and / or pointing device. Devices connected to input / output subsystem 1160 can be used to provide information to processors 1140 by a user of classification system 1100. Devices connected to input / output subsystem 1160 can include network or communication controllers, input devices (keyboard, mouse, touchscreen, etc.) and output devices (printer or display). Input / output subsystem 1160 can also provide an interface to devices such as a display and / or printer that can communicate results of operations of classification system 1100 to the display and / or printer and thus to a user of classification system 1100.

[0185] An optional upgrade loop 1180 can be included as an interface for providing upgrades to the classification system 1100. The upgrades can include upgrades to code executed by the processor 1140 stored in the memory 1150 and / or the storage loop 1170. The upgrades provided through the upgrade loop 1180 can also include upgrades to portions of the storage loop 1170 related to databases and / or other data storage forms that maintain information for the classification system 1100 such as identification signatures, weights, thresholds, and the like.

[0186] The sample input loop 1110 of the classification system 1100 can provide an interface for a platform as described above to receive biological samples to be analyzed. The sample input loop 1110 can include mechanical and electrical elements that receive biological samples provided by a user to the classification system 1100 and transport the biological samples to be processed within the classification system 1100 and / or the platform. The sample input loop 1110 can include a barcode reader to identify barcoded containers identifying samples and / or assay formats. The sample processing loop 1120 can also process the biological samples within the classification system 1100 and / or the platform to prepare the biological samples for automated analysis. The sample analysis loop 1130 can automatically analyze the processed biological samples. The sample analysis loop 1130 can be used to measure gene expression levels of a predefined set of genes using the biological samples provided to the classification system 1100. The sample analysis loop 1130 can also generate normalized gene expression values by normalizing the gene expression levels. The sample analysis loop 1130 can retrieve from the storage loop 1170 a bacterial acute respiratory infection (ARI) classifier, a viral ARI classifier, and a non-infectious disease classifier, the one or more classifiers including predefined weighting values (i.e., coefficients) for each of the genes of the predefined set of genes. The sample analysis loop 1130 can input the normalized gene expression values into one or more acute respiratory disease classifiers selected from the bacterial ARI classifier, the viral ARI classifier, and the non-infectious disease classifier. The sample analysis loop 1130 can calculate a probability of etiology for one or more of a bacterial ARI, a viral ARI, and a non-infectious disease based on the one or more classifiers and control outputting a determination of whether the acute respiratory disease in the subject is of a bacterial origin, a viral origin, a non-infectious origin, or some combination thereof through the input / output loop 1160.

[0187] The sample input circuit 1110, sample processing circuit 1120, sample analysis circuit 1130, input / output circuit 1160, storage circuit 1170, and / or upgrade circuit 1180 can operate at least partially under the control of one or more processors 1140 of the classification system 1100. As used herein, operating "under the control of" the processor 1140 means that operations by the sample input circuit 1110, sample processing circuit 1120, sample analysis circuit 1130, input / output circuit 1160, storage circuit 1170, and / or upgrade circuit 1180 can be at least partially performed and / or directed by the processor 1140, but does not exclude at least a portion of the operations of those components that are solely electrically or mechanically automated. As described herein, the processor 1140 can control the operation of the classification system 1100 by executing computer program code.

[0188] Computer program code for carrying out operations for aspects of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, or the like, conventional procedural programming languages, such as the "C" programming language, Visual Basic, Fortran 2003, Perl, COBOL 2002, PHP, ABAP, dynamic programming languages such as Python, Ruby and Groovy, or other programming languages. The program code can execute entirely on the classification system 1100, partly on the classification system 1100, as a stand-alone software package, partly on the classification system 1100 and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the classification system 1100 through any type of network, including a local area network (LAN) or a wide area network (WAN), or the

[0189] In some embodiments, the system includes computer readable code that can convert quantitative or semi-quantitative gene expression detection into a cumulative score or probability of an etiology of an ARI.

[0190] In some embodiments, the system is a sample-to-result system, and the elements are integrated such that a user can simply insert a biological sample to be tested, and receive a result output from the system after some time, preferably a short amount of time, e.g., 30 or 45 minutes, or 1, 2, or 3 hours, up to 8, 12, 24, or 48 hours.

[0191] It is to be understood that the application in its application is not limited to the details of construction and the arrangement of components set forth in the following description or illustrated in the following drawings. The application is capable of other embodiments and of being practiced or being carried out in various ways.

[0192] Unless otherwise indicated herein, recitation of a range of values herein is merely intended to serve as a shorthand method of referring individually to each separate value falling within the range. Unless otherwise indicated herein, each individual value is incorporated into the specification as if it were individually recited herein. All methods described herein can be performed in any suitable order unless otherwise indicated herein. The use of any and all examples, or exemplary language (e.g., "such as") provided herein, is intended merely to better illuminate the application and does not pose a limitation on the scope of the application unless otherwise claimed. No language in the specification should be construed as indicating any non-claimed element as essential.

[0193] It is also to be understood that the description herein of any range of values is inclusive of all values from the lower value to the higher value. For example, if a concentration range is stated to be 1% to 50%, it is intended that values such as 2% to 40%, 10% to 30%, or 1% to 3%, etc., are expressly enumerated in this specification. These are only examples of what is specifically intended, and all possible combinations of numerical values between the lowest value and the highest value enumerated, and including the lowest and highest values, are to be considered to be expressly stated in this application.

[0194] The following examples are illustrative only and are not intended to limit the scope. Example

[0195] Example 1. Host gene expression classifier diagnoses acute respiratory disease etiology

[0196] Acute respiratory infections due to bacterial or viral pathogens are one of the most common reasons for seeking medical care. Current pathogen-based diagnostic methods are unreliable or not timely, so most patients receive inappropriate antibiotics. Host response biomarkers provide an alternative diagnostic approach to direct antimicrobial use.

[0197] We asked whether host gene expression patterns in an acute care setting would distinguish infectious from non-infectious etiologies. Among those with acute respiratory infections, we determined whether infectious disease was due to viral or bacterial pathogens.

[0198] The sample constituting the basis for the findings was selected from an observational cohort study conducted at four tertiary care hospital emergency departments and student health facilities. From a larger cohort of patients with suspected sepsis, 44 healthy controls and 273 patients with community-onset acute respiratory infection or non-infectious disease were selected (CAPSOD study). The mean age was 45 years and 45% of the participants were male. Further demographic information can be found in Table 1 of Tsalik et al. (2016) Sci Transl Med 9(322): 1-9, which is incorporated herein by reference.

[0199] Clinical phenotypes were determined by manual chart review. Routine microbiology testing and multiplex PCR for respiratory viral pathogens were performed. Peripheral whole blood gene expression was measured using microarrays. Sparse logistic regression was used to develop classifiers for bacterial versus viral versus non-infectious disease. Five independently obtained datasets including 328 individuals were used for validation.

[0200] Gene expression-based classifiers were developed for bacterial acute respiratory infection (71 probes), viral acute respiratory infection (33 probes), or non-infectious etiology (26 probes). The three classifiers were applied to 273 patients, with category assignment determined by highest predicted probability. Overall accuracy was 87% (238 / 273 in agreement with clinical determination), which was more accurate than procalcitonin (78%, p<0.03) and three published classifiers for bacterial versus viral infection (78-83%). The classifiers developed here were externally validated (AUC 0.90-0.99) in five publicly available datasets. We compared the classification accuracy of the host gene expression-based assays to procalcitonin and diagnosis by clinical determination, which included bacterial or viral acute respiratory infection or non-infectious disease.

[0201] Peripheral blood gene expression responses of the host to infection provide complementary diagnostic strategies to those already in use. 8 This strategy has successfully characterized host responses to viral 8-13 and bacterial ARI 11,14 Despite these advances, multiple issues preclude their use as diagnostics in the patient care setting. An important consideration in developing host-based molecular recognition signatures is that they be formed in the population for which use is intended. 15However, almost all published gene expression-based ARI classifiers use healthy individuals as controls and focus on small or homogeneous populations and thus are not optimized for the acute care setting where patients present with undifferentiated symptoms. In addition, the statistical methods used to identify the gene-expression classifiers typically include redundant genes based on clustering, univariate testing, or pathway association. These strategies identify relevant biology but do not maximize diagnostic performance. An alternative as exemplified herein is to combine genes from non-relevant pathways to generate classifiers that provide more information.

[0202] Methods

[0203] Classifier derivation cohort

[0204] The study was approved by the relevant ethics review board and was in compliance with the Declaration of Helsinki. All subjects or their legally authorized representatives provided written informed consent.

[0205] Patients with community-onset, suspected infection were enrolled in the emergency department at Duke University Medical Center (DUMC; Durham, NC), Durham VA Medical Center (DVAMC; Durham, NC), or Henry Ford Hospital (Detroit, MI) as part of the Community Acquired Pneumonia & Sepsis Outcome Diagnosis Study (Clinical Trial Identifier NCT00258869). 16-19 Additional patients were enrolled at the UNC Health Care Emergency Department (UNC; Chapel Hill, NC) as part of the Community Acquired Pneumonia and Sepsis Study. Patients were eligible if they had a known or suspected infection and if they exhibited two or more systemic inflammatory response syndrome (SIRS) criteria. 20 ARI cases included patients with upper or lower respiratory symptoms as determined by emergency medicine (SWG, EBQ) or infectious disease (ELT) physicians. Determination was based on retrospective, manual chart review performed at least 28 days after enrollment and prior to any gene expression-based classification using previously published criteria. 17All of the information used to support these determinations would not have been available to the clinician at the time of their assessment. Seventy patients with microbiologically confirmed bacterial ARI were identified, including four with pharyngitis and 66 with pneumonia. Microbial etiology was determined using routine culture of blood or respiratory samples, urine antigen tests (Streptococcus or Legionella), or serological testing (Mycoplasma). Patients with viral ARI (n=115) were determined based on identification of viral etiology and compatible symptoms. In addition, 48 Duke University students with confirmed viral ARI using the same determination methods as part of the DARPA Predicting Health and Disease study. The ResPlex II v 2.0 viral PCR multiplex assay (Qiagen; Hilden, Germany) augmented the clinical testing for viral etiology identification. This panel detects influenza A and B, adenovirus (B, E), parainfluenza viruses 1-4, respiratory syncytial viruses A and B, human metapneumovirus, human rhinovirus, coronaviruses (229E, OC43, NL63, HKU1), coxsackie / echo viruses, and bocavirus. At the time of determination, a subset of the enrolled patients were determined to have non-infectious disease (n=88) (Table 8). The determination of "non-infectious disease" was made only when an alternative diagnosis was established and the results of any routinely ordered microbiological testing could not support an infectious etiology. Finally, healthy controls (n=44; median age 30 years; range 23-59) were enrolled as part of a study on the effects of aspirin on platelet function in healthy volunteers without symptoms, where gene expression analysis was performed at pre-aspirin challenge time points. 21

[0206] Procalcitonin measurement

[0207] Concentrations were measured at different stages during the study and, as a consequence, different platforms were used based on availability. Some serum measurements were performed by electrochemiluminescence immunoassay on a Roche Elecsys 2010 analyzer (Roche Diagnostics, Laval, Canada). Additional serum measurements were performed using miniVIDAS immunoassay (bioMerieux, Durham NC, USA). When serum was not available, measurements were performed by immunofluorescence in plasma-EDTA by Phadia Immunology Reference Laboratory using B•R•A•H•M•SPCT sensitive KRYPTOR (Thermo Fisher Scientific, Portage MI, USA). Replicates were performed for some paired serum and plasma samples, showing equal concentrations. Therefore, all procalcitonin measurements were treated equally without considering the testing platform.

[0208] Microarray generation

[0209] Patients were recruited at initial clinical presentation and samples were collected for analysis. After the determination as described above, 317 subjects with clear clinical phenotype were selected for gene expression analysis. Total RNA was extracted from human blood using PAXgene Blood RNA Kit (Qiagen, Valencia, CA) according to the manufacturer's protocol. RNA quantity and quality were assessed using a Nanodrop spectrophotometer (Thermo Scientific, Waltham, MA) and an Agilent 2100 Bioanalyzer (Agilent, Santa Clara, CA), respectively. Microarray RMA normalization. Hybridization and data collection were performed according to the Affymetrix technical manual using GeneChip Human Genome U133A 2.0 arrays (Affymetrix, Santa Clara, CA) at Expression Analysis (Durham, NC).

[0210] Statistical analysis

[0211] The transcriptomes of 317 subjects (273 patients with disease and 44 healthy volunteers) were measured in two microarray batches with seven overlapping samples (GSE63990). Exploratory principal components analysis and hierarchical clustering revealed substantial batch differences. These were corrected by first estimating and removing the per-probe mean batch effect using a Bayesian fixed effects model. Next, we fitted a robust linear regression model using the seven overlapping samples with Huber loss function, which was used to adjust the remaining expression values.

[0212] While reducing the risk of overfitting, sparse classification methods like sparse logistic regression simultaneously perform classification and variable selection. 21 Thus, separate gene selection strategies like univariate tests or sparse factor models are not necessary. Here, after batch correction, we fitted a sparse logistic regression model independently with the 40% probesets with the largest variance to each of the binary tasks. 22 Specifically, we used Lasso-regularized generalized linear models with binomial likelihood to select the regularization parameter using nested cross-validation. The code uses the Glmnet toolbox in Matlab. This generated bacterial ARI, viral ARI, and non-infectious disease classifiers. Assuming that each binary classifier estimates the class membership probabilities (e.g., the probability of being bacterial versus viral or non-infectious in the case of the bacterial ARI classifier), we can combine the three classifiers into a single decision model (referred to as the ARI classifier) following the one-out-of-all scheme where the class label is assigned where the largest membership probability is. 21 Metrics of classification performance include the area-under-the-receiving-operating-characteristic-curve (AUC) for binary outcomes and the confusion matrix for ternary outcomes. 23

[0213] Validation

[0214] ARI classifiers were validated in the same population from which they came using leave-one-out cross-validation. External validation was performed using publicly available human gene expression datasets (GSE6269, GSE42026, GSE40396, GSE20346, and GSE42834) from 328 individuals. Datasets were selected if they included at least two clinical groups (bacterial ARI, viral ARI, or non-infectious disease). To match probes between different microarray platforms, each ARI classifier probe was converted to gene symbol, which was used to identify the corresponding target microarray probe.

[0215] Results

[0216] Bacterial ARI, viral ARI, and non-infectious disease classifiers

[0217] In the process of generating host gene expression-based classifiers that distinguish clinical states, all relevant clinical phenotypes should be represented during the model training process. This confers specificity, allowing the model to be applied to these included clinical groups but not to clinical phenotypes that were not present in the model training. 15 The target population for ARI diagnosis includes not only patients with viral and bacterial etiologies, but must also be distinguished from the comparative scenario, those without bacterial or viral ARI. In the past, healthy individuals served as the non-infectious control group. However, this fails to account for how patients with non-infectious diseases that can present with similar clinical symptoms would be classified, serving as a potential source of diagnostic error. To our knowledge, no gene expression-based classifier for ARI has included sick, non-infectious controls in its derivation. We therefore recruited a large, heterogeneous population with initial clinical presentation of either community-onset viral ARI (n=115), bacterial ARI (n=70), or non-infectious disease (n=88) (Table 8). We also included a cohort of healthy adult controls (n=44) to define the most appropriate control population for ARI classifier development.

[0218] We first determined whether gene expression classifiers derived using healthy individuals as controls could accurately classify patients with non-infectious disease. Gene expression classifiers for these conditions were generated using array data from patients with bacterial ARI, viral ARI, and healthy controls. Leave-one-out cross-validation showed highly accurate discrimination between bacterial ARI (AUC 0.96), viral ARI (AUC 0.95), and healthy (AUC 1.0) subjects with 90% combined accuracy (Figure 1). The bacterial ARI classifier was then applied to the non-infectious disease cohort. The classifier showed 87% accuracy in classifying patients with non-infectious disease (Figure 2). The viral ARI classifier showed 88% accuracy in classifying patients with viral ARI (Figure 3). The healthy classifier showed 100% accuracy in classifying healthy controls (Figure 4). Figure 7). However, when the classifiers were applied to the sick - non-infectious - patients, 48 / 88 were identified as bacterial, 35 / 88 as viral, and 5 / 88 as healthy. This highlights that healthy individuals are poor surrogates for patients with non-infectious diseases in the biomarker discovery process.

[0219] Therefore, we used non-infectious disease controls instead of healthy controls to derive the ARI classifiers. Specifically, three gene expression classifiers for bacterial ARI, viral ARI, and host response to non-infectious disease were generated using array data from these three groups ( Figure 5 ). Specifically, the bacterial ARI classifier was run to identify those patients with bacterial ARI relative to viral ARI or non-infectious disease positive. The viral ARI classifier was run to identify those patients with viral ARI relative to bacterial ARI or non-infectious disease positive. A non-infectious disease classifier was not generated to be used for positive identification of all non-infectious diseases, which would have required a sufficient description of all such cases.

[0220] Instead, it was generated as an alternative category so that patients not having bacterial or viral ARI could be assigned accordingly. Furthermore, we assumed that such sick but non-infectious patients are more clinically relevant controls, as healthy people are unlikely to be the target of such a classification task.

[0221] Six statistical strategies were employed to generate these gene expression classifiers: linear support vector machine, supervised factor model, sparse multinomial logistic regression, elastic net, K- nearest neighbors, and random forest. All were performed similarly, although sparse logistic regression required the least number of classifier genes and was superior to the other strategies due to small tolerance (data not shown). We also compared the strategy of generating three separate binary classifiers to a single multinomial classifier that would assign one of the three clinical categories to a given subject at the same time. This latter approach required more genes and resulted in poorer accuracy. Therefore, we applied the sparse logistic regression model to define bacterial ARI, viral ARI, and non-infectious disease classifiers containing 71, 33, and 26 probe identifiers, respectively. The probes and classifier weights are shown in Table 9.

[0222] Clinical decisions are rarely binary, requiring the simultaneous discrimination of multiple diagnostic possibilities. We applied all three classifiers collectively defined as ARI classifiers to assign probabilities of bacterial ARI, viral ARI, and non-infectious disease using leave-one-out cross-validation ( Figure 6). These conditions are not mutually exclusive. For example, the presence of bacterial ARI does not preclude the concurrent presence of viral ARI or non-infectious illness. Furthermore, the assigned probabilities represent the degree to which the patient's gene expression response matches the typical identifying signatures of that condition. Because each identifying signature is intended to function independently of the others, the probabilities are not expected to sum to one. To simplify classification, the highest predicted probability determines the category assignment. Overall classification accuracy was 87% (238 / 273 in agreement with the determined phenotype).

[0223] Bacterial ARI was confirmed in 58 / 70 (83%) patients and excluded 179 / 191 (94%) patients without bacterial infection. Viral ARI was confirmed in 90% (104 / 115) and excluded in 92% (145 / 158) cases. Using the non-infectious illness classifier, infection was excluded in 86% of cases (76 / 88). Assuming that incidence can vary for many reasons, including infection type, patient characteristics, or location, sensitivity analysis was performed on the positive and negative predictive values for all three classifiers (Table 2). Figure 8 ) For both bacterial and viral classifications, the predictive values remained high across the range of incidence estimates, including those typically found for ARI.

[0224] To determine if there was any age effect, we included it as a variable in the classification scheme. This resulted in two additional correct classifications, possibly attributable to over-representation of younger individuals in the viral ARI group. However, we did not observe a statistically significant difference attributable to age between correctly and incorrectly classified subjects (Wilcoxon rank sum p = 0.17).

[0225] We compared this performance to procalcitonin, a widely used biomarker specific for bacterial infection. Procalcitonin concentrations were determined for 238 subjects where samples were available, and this subgroup was compared to ARI classifier performance. Procalcitonin concentrations > 0.25 μg / L assigned patients as having a bacterial ARI, while values < 0.25 μg / L assigned patients as non-bacterial, which could be viral ARI or non-infectious illness. Procalcitonin correctly classified 186 / 238 (78%) of patients compared to 204 / 238 (86%) using the ARI classifier (p = 0.03). However, the accuracy of both strategies varied depending on the classification task. For example, in discriminating viral from bacterial ARIs, the performance was similar. Procalcitonin correctly classified 136 / 155 (AUC 0.89) compared to 140 / 155 of the ARI classifier (p-value = 0.65 using McNemar's test with Yates correction). However, the ARI classifier was significantly better than procalcitonin in discriminating bacterial ARIs from non-infectious illness [105 / 124 versus 79 / 124 (AUC 0.72); p-value < 0.001] and in discriminating bacterial ARIs from all other etiologies including viral and non-infectious etiologies [215 / 238 versus 186 / 238 (AUC 0.82); p-value = 0.02].

[0226] We next compared the ARI classifier to three published gene expression classifiers for bacterial versus viral infection, each of which was derived without non-infectious illness controls. These included a 35 probe classifier derived from children with influenza or bacterial sepsis (Ramilo) 11 ; a 33 probe classifier derived from children with febrile viral illness or bacterial infection (Hu) 14 ; and a 29 probe classifier derived from adult ICU patients with community-acquired pneumonia or influenza (Parnell) 12We hypothesized that a classifier generated using only patients with viral or bacterial infections would perform poorly when applied to a clinically relevant population that includes patients with disease but not infectious disease. Specifically, when given an individual who has neither a bacterial infection nor a viral infection, the previously disclosed classifiers would not accurately assign the third, alternative category to the individual. We therefore applied the resulting and disclosed classifiers to our cohort of 273 patients. The resulting ARI classifier was better at distinguishing between bacterial ARI, viral ARI, and non-infectious disease (Yates-corrected McNemar’s test, p = 0.002 vs. Ramilo; p = 0.0001 vs. Parnell; and p = 0.08 vs. Hu) (Table 6). 24,25 This underscores the importance of deriving the gene expression classifier in a cohort that represents the intended use population of patients with ARI, which should include non-infectious disease. 15

[0227] Inconsistent classifications

[0228] To better understand the ARI classifier performance, we reviewed the 35 cases of inconsistency individually. Nine adjudicated bacterial infections were classified as viral and three were classified as non-infectious disease. Four viral infections were classified as bacterial and seven were classified as non-infectious. Eight non-infectious cases were classified as bacterial and four were classified as viral. We did not observe a consistent pattern in the cases of inconsistency, however, notable examples included atypical bacterial infections. One patient with Mycoplasma pneumoniae and one of three patients with Legionella pneumonia based on seroconversion were classified as viral ARI. Of the six patients with non-infectious disease due to autoimmune or inflammatory disease, only one patient adjudicated to have Still’s disease was classified as having a bacterial infection. See also Table 3 of Tsalik et al. (2016) Sci Transl Med 9(322): 1-9, which is incorporated herein by reference.

[0229] External validation

[0230] Generating classifiers from large size, gene expression data can result in overfitting. We therefore validated the ARI classifiers in silico using gene expression data from 328 individuals represented in five available datasets (GSE6269, GSE42026, GSE40396, GSE20346, and GSE42834). These were chosen because they included at least two relevant clinical groups, differing in age, geographic distribution, and disease severity (Table 7). The ARI classifiers were applied to four datasets with both bacterial and viral ARI, with AUC ranging from 0.90-0.99. Finally, GSE42834 included patients with bacterial pneumonia (n=19), lung cancer (n=16), and sarcoidosis (n=68). Overall classification accuracy was 96% (99 / 103), corresponding to an AUC of 0.99. GSE42834 included five subjects with bacterial pneumonia before and after treatment. All five subjects showed a treatment-dependent resolution of bacterial infection. See also Tsalik et al. (2016) Sci Transl Med 8(322): 1-9, which is incorporated herein by reference. Figures 3-8

[0231] Biological pathways

[0232] If no additional diagnostic values are present, the sparse logistic regression model generating the classifiers makes the selection of genes from a given pathway disadvantageous. Therefore, regular gene enrichment pathway analysis cannot be properly performed. Moreover, such regular gene enrichment analysis has been described. 9,12,14,28,29 Instead, a literature review was performed for all classifier genes (Table 10). Overlaps between the bacterial, viral, and non-infectious disease classifiers are shown in Figure 9

[0233] The viral classifier includes known categories of anti-viral responses, such as interferon response, T cell signaling, and RNA processing. The viral classifier has the greatest representation of RNA processing pathways such as KPNB1, which is involved in nuclear transport and is taken up by viruses for transport of viral proteins and genomes. 26,27 Downregulation of which suggests that it can play an anti-viral role in the host response.

[0234] The bacterial classifier includes a broad range of cellular processes, especially cell cycle regulation, cell growth, and differentiation. The bacterial classifier includes genes important in T, B, and NK cell signaling. Unique to the bacterial classifier are genes involved in oxidative stress as well as fatty acid and amino acid metabolism, consistent with the metabolic chaos associated with sepsis. 28

[0235] Overview of clinical applicability ​​

[0236] We determined that host gene expression changes are particularly specific to the class of annoying pathogens and can be used to discriminate the etiology of common respiratory illnesses. This creates the opportunity to develop and use gene expression classifiers as a novel diagnostic platform to fight inappropriate antibiotic use and emerging antibiotic resistance. Using sparse logistic regression, we developed host gene expression profiles that accurately distinguish between bacterial and viral etiology in patients with acute respiratory symptoms (external validation AUC 0.90-0.99). The ARI classifier confers a high negative predictive value with non-infectious disease controls within a broad range of prevalence estimates.

[0237] Respiratory infections caused 3.2 million deaths and 164 million disability-adjusted life years lost in 2011 worldwide, more than any other cause. 1,2 Despite being viral etiology in most cases, 73% of ambulatory patients with acute respiratory infections (ARI) in the US are prescribed antibiotics, accounting for 41% of all antibiotic prescriptions in this context. 3,4 Even when the viral etiology is microbiologically confirmed, this does not rule out possible concurrent bacterial infection, leading to antimicrobial prescription "just in case". This empiricism contributes to antimicrobial resistance 5,6 which is considered a national security priority. 7 The encouraging indicators provided in this example provide the opportunity to provide clinically actionable results that will optimize treatment and mitigate emerging antibiotic resistance.

[0238] Many studies have made significant progress in developing host response diagnostics for ARI. This includes responses to respiratory viruses 8,10-12,14 bacterial etiology in ICU populations 12,30 and tuberculosis 31-33 . Typically, these define host response profiles compared to health status, providing valuable insights into host biology. 16,34,35 However, these listed genes are not optimal for diagnostic applications because the gene expression profiles as elements of the diagnosis do not represent the population that will be applied to the test. 15 Healthy individuals do not have acute respiratory illness, so they are excluded from the development of host response diagnostics reported in this document.

[0239] Including patients with bacterial and viral infections allows for the differentiation of these two states, but does not address how to classify non-infectious diseases. It is important to include this phenotype because patients exhibit infectious and non-infectious etiologies that can share symptoms. That is, symptoms can not provide a high degree of diagnostic certainty to the clinician. The current approach, which is aware of the need to include the three most likely states for ARI symptoms, can be applied to the undifferentiated clinical population where such testing is in the greatest demand.

[0240] The few instances of inconsistent classification that arise can result from errors in the classification or clinical phenotype determination. Errors in the clinical phenotype determination can be attributed to the inability to identify the causative pathogen due to the limitations of current microbiology diagnostics. Alternatively, some non-infectious disease processes can in fact be infection-related according to mechanisms that have not yet been discovered. The instances of inconsistency are not clearly explained by uniform variables such as pathogen type, syndrome, or patient characteristics. Therefore, the gene expression classifiers presented herein can be influenced by other factors, including patient-specific variables (e.g., treatment, comorbidity, duration of illness); test-specific variables (e.g., sample handling, assay conditions, RNA quality and yield); or variables that have not yet been elucidated.

[0241] Example 2: Classification performance in patients with superinfection defined by discrimination of bacterial and viral pathogens

[0242] In addition to determining that age did not significantly affect classification accuracy, we also evaluated whether the severity of illness or the etiology of SIRS affected classification. Patients with viral ARI tend to be less ill as evidenced by lower rates of hospitalization. In multiple cohorts, hospitalization was used as a marker of disease severity and its effect on classification performance was evaluated. This test did not show a difference (Fisher's exact test, p-value of 1). In addition, the SIRS control cohort included subjects with both respiratory and non-respiratory etiologies. We evaluated whether classification was different in subjects with respiratory SIRS versus non-respiratory SIRS and determined that it was not (Fisher's exact test, p-value of 0.1305).

[0243] Some patients with ARI will have both bacterial and viral pathogens identified, often referred to as co-infection. However, it is unclear how the host responds in such cases. At different times in the patient's clinical course, the illness can be caused by bacteria, virus, both, or neither. We therefore determined how to implement the bacterial and viral ARI classifiers in a population with bacterial and viral co-identification. GSE60244 included bacterial pneumonia (n=22), viral respiratory infection (n=71), and bacterial / viral co-identification (n=25). The co-identified group was defined by the presence of both bacterial and viral pathogens, without further subclassification into the likelihood of bacterial or viral disease. We trained the classifiers on subjects in GSE60244 with bacterial or viral infection and then validated on those with co-identification Figure 10 ). The host response was considered positive above a probability threshold of 0.5. We observed all four possible categories. Six of 25 subjects had a positive bacterial signature; 14 / 25 had a viral response; 3 / 25 had a positive bacterial and viral signature; and 2 / 25 had neither.

[0244] The primary clinical decision facing the clinician is whether to prescribe an antibacterial agent. A simpler diagnostic strategy can focus only on the probability of bacterial ARI according to the results from the bacterial ARI classifier. However, it is valuable to provide information about viral or non-infectious alternatives. For example, a high probability of an alternative diagnosis can increase confidence in not using an antibacterial agent in a patient with a low probability of bacterial ARI. In addition, a complete diagnostic report can identify concurrent illnesses that a single classifier would miss. We observed this when validating the bacterial and viral co-identification in a population. These patients are more commonly referred to as "co-infected." To have an infection, there must be a pathogen, a host, and an ill-adapted interaction between the two. Identifying only bacterial and viral pathogens should not imply co-infection. Although we cannot know the true infection status in the 25 subjects tested with evidence of bacterial / viral co-identification, the host response classifier indicates the existence of multiple host response states. Figure 10 is an information graphic of the infection status that can be used by a clinician to diagnose the etiology of ARI.

[0245] References

[0246] 1. Organization. WH. Global health estimates summary tables: Deaths by cause, age and sex. 2013. Accessed May 14, 2014.

[0247] 2. Organization. WH. Global Health Estimates Summary Tables: DALYs by cause, age and sex. 2013. Accessed May 14, 2014.

[0248] 3. Shapiro et al. Antibiotic prescribing for adults in ambulatory care in the USA, 2007-09. J Antimicrob Chemother. 2014;69(1):234-240.

[0249] 4. Lee et al. Outpatient antibiotic prescribing in the United States: 2000 to 2010. BMC Medicine. 2014;12:96.

[0250] 5. Gould. Antibiotic resistance: the perfect storm. Int J Antimicrob Ag. 2009;34, Supplement 3(0):S2-S5.

[0251] 6. Kim and Gallis. Observations on spiraling empiricism: its causes, allure, and perils, with particular reference to antibiotic therapy. Am J Med. 1989;87(2):201-206.

[0252] 7. Obama B. Executive Order - Combating Antibiotic-Resistant Bacteria. In: House TW, ed: Office of the Press Secretary; 2014.

[0253] 8. Zaas et al. Gene expression signatures diagnose influenza and other symptomatic respiratory viral infections in humans. Cell host & microbe. 2009; 6(3):207-217.

[0254] 9. Woods et al. A Host Transcriptional Signature for Presymptomatic Detection of Infection in Humans Exposed to Influenza H1N1 or H3N2. PLoS ONE. 2013; 8(1):e52198.

[0255] 10. Mejias et al. Whole Blood Gene Expression Profiles to Assess Pathogenesis and Disease Severity in Infants with Respiratory Syncytial Virus Infection. PLoS medicine. 2013; 10(11):e1001549.

[0256] 11. Ramilo et al. Gene expression patterns in blood leukocytes discriminate patients with acute infections. Blood. 2007; 109(5):2066-2077.

[0257] 12. Parnell et al. A distinct influenza infection signature in the blood transcriptome of patients who presented with severe community acquired pneumonia. Crit Care. 2012; 16(4):R157.

[0258] 13. Zaas et al. A host-based RT-PCR gene expression signature to identify acute respiratory viral infection. Sci trans med. 2013;5(203):203ral26.

[0259] 14. Hu et al. Gene expression profiles in febrile children with defined viral and bacterial infection. Proc Natl Acad Sci U S A. 2013.

[0260] 15. Lytkin et al. Expanding the understanding of biases in development of clinical-grade molecular signatures: a case study in acute respiratory viral infections. PLoS One. 2011;6(6):e20662.

[0261] 16. Ahn et al. Gene expression-based classifiers identify Staphylococcus aureus infection in mice and humans. PLoS One. 2013;8(1):e48979.

[0262] 17. Glickman et al. Disease progression in hemodynamically stable patients presenting to the emergency department with sepsis. Acad Emerg Med. 2010;17(4):383-390.

[0263] 18. Tsalik et al. Discriminative value of inflammatory biomarkers for suspected sepsis. J Emerg Med. 2012;43(1):97-106.

[0264] 19. Tsalik et al. Multiplex PCR to diagnose bloodstream infections in patients admitted from the emergency department with sepsis. J Clin Microbiol. 2010;48(1):26-33.

[0265] 20. Bone et al. Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. The ACCP / SCCM Consensus Conference Committee. American College of Chest Physicians / Society of Critical Care Medicine. Chest. 1992;101(6): 1644-1655.

[0266] 21. Bishop. Pattern Recognition and Machine Learning (Information Science and Statistics). Springer-Verlag New York, Inc.; 2006.

[0267] 22. Friedman et al. Regularization Paths for Generalized Linear Models via Coordinate Descent. J. Stat. Softw. 2010;33(1): 1-22.

[0268] 23. Fawcett. An introduction to ROC analysis. Pattern Recognition Letters. 2006;27(8):861-874.

[0269] 24. McNemar. Note on the sampling error of the difference between correlated proportions or percentages. Psychometrika. 1947;12(2):153-157.

[0270] 25. Yates. Contingency Tables Involving Small Numbers and the χ2 Test. Supplement to the Journal of the Royal Statistical Society. 1934;1(2):217-235.

[0271] 26. Bukrinsky et al. Active nuclear import of human immunodeficiency virus type 1 preintegration complexes. Proc Natl Acad Sci U S A. 1992;89(14):6580-6584.

[0272] 27. Ghildyal et al. Nuclear import of the respiratory syncytial virus matrix protein is mediated by importin beta1 independent of importin alpha. Biochemistry. 2005;44(38):12887-12895.

[0273] 28. Langley et al. An Integrated Clinico-Metabolomic Model Improves Prediction of Death in Sepsis. Sci Trans Med. 2013;5(195):195ra195.

[0274] 29. Schappert and Rechisteiner. Ambulatory medical care utilization estimates for 2007. In: Stat VH, ed. Vol 132 011.

[0275] 30. Severino et al. Patterns of Gene Expression in Peripheral Blood Mononuclear Cells and Outcomes from Patients with Sepsis Secondary to Community Acquired Pneumonia. PLoS ONE. 2014;9(3):e91886.

[0276] 31. Anderson et al. Diagnosis of Childhood Tuberculosis and Host RNA Expression in Africa. N Engl J Med. 2014;370(18): 1712-1723.

[0277] 32. Berry et al. An interferon-inducible neutrophil-driven blood transcriptional signature in human tuberculosis. Nature. 2010;466(7309):973-977.

[0278] 33. Kaforou et al. Detection of tuberculosis in HIV-infected and -uninfected African adults using whole blood RNA expression signatures: a case-control study. PLoS medicine. 2013;10(10):el001538.

[0279] 34. Banchereau et al. Host immune transcriptional profiles reflect the variability in clinical disease manifestations in patients with Staphylococcus aureus infections. PLoS One. 2012; 7(4):e34390.

[0280] 35. Herberg et al. Transcriptomic Profiling in Childhood H1N1 / 09 Influenza Reveals Reduced Expression of Protein Synthesis Genes. J Infect Dis. 2013; 208(10): 1664-1668.

[0281] 36. Bloom et al. Transcriptional blood signatures distinguish pulmonary tuberculosis, pulmonary sarcoidosis, pneumonias and lung cancers. PLoS One. 2013; 8(8):e70630.

[0282] Table 6: Performance characteristics of ARI classifiers derived. The combined bacterial ARI, viral ARI, and non-infectious disease classifiers were validated in a population of bacterial ARI (n=70), viral ARI (n=115), or non-infectious disease (NI, n=88) using leave-one-out cross-validation. The three published bacterial versus viral classifiers were identified and applied to this same population as a comparator. Data are given as numbers (%). Shaded cells indicate correct classification.

[0283]

[0284] Table 7: External validation of ARI classifiers (combined bacterial ARI, viral ARI, and non-infectious classifiers). Five gene expression Omnibus datasets were identified based on inclusion of at least two of the relevant clinical groups: viral ARI, bacterial ARI, non-infectious disease (NI).

[0285]

[0286]

[0287]

[0288]

[0289] a This patient was adjudicated to have a bacterial ARI, and Bacillus sp. was identified as the etiologic agent. We later realized that, despite the clinical history being consistent with bacterial pneumonia, Bacillus sp. was not the correct microbiological etiology. Because this error was identified after model derivation, we included this subject in all subsequent analyses.

[0290]

[0291]

[0292]

[0293]

[0294]

[0295] Table 10. Genes in the bacterial ARI, viral ARI, and non-infectious disease (NI) classifiers grouped by biological process. Gene accession numbers are provided in Table 9.

[0296]

[0297] * Genes listed in more than one classifier. In cases where such overlapping genes have different expression directions, increased expression is indicated by (+) and decreased expression by (-).

[0298] Example 3: Bacterial / viral / SIRS assay considered on TLDA platform

[0299] We will develop custom multi-analyte, quantitative real-time PCR (RT-PCR) assays on the 384-well TaqMan Low Density Array (TLDA, Applied Biosystems) platform. TLDA cards will be manufactured with one or more TaqMan primer / probe sets specific for mRNA transcripts of one or more classifier genes in each well, along with multiple endogenous control RNA targets (primer / probe sets) for data normalization. For each patient sample, purified total RNA will be reverse transcribed to cDNA, loaded into a master well and partitioned into each assay well via microfluidic channels by centrifugation. TaqMan hydrolysis probes rely on 5' to 3' exonuclease activity to cleave the dual-labeled probe during hybridization to the complementary target sequence, and each round of amplification results in a fluorescent signal. In this way, "real-time" quantitative detection of accumulating PCR products is possible. During exponential amplification and detection, the number of PCR cycles in which the fluorescent signal crosses a detection threshold is the threshold cycle (C t ) or quantification cycle (C q ) as determined by commercial software for RT-PCR instruments. To quantify gene expression, the C t of the target RNA is subtracted from the C t of the endogenous normalization RNA (or the geometric mean of multiple normalization RNAs), resulting in a ΔC t value for each RNA target within a sample, which represents the relative expression of the target RNA normalized for variability in the amount or quality of input sample RNA or cDNA.

[0300] Quantified gene signature data are then processed using a computer and according to the probit classifier described above (Equation 1) and reproduced here. The normalized gene expression level of each gene in the signature is the explanatory or independent variable or feature used in the classifier, and in this example, the general form of the classifier is a probit regression equation:

[0301]

[0302] where the condition is bacterial ARI, viral ARI, or non-communicable disease; Φ(.) is the probit link function; {β1, β2,..., β d} are coefficients obtained during training; {X1, X2,..., X d} is the normalized gene expression value for the signature; and d is the size of the signature (number of genes). The value of the coefficient for each explanatory variable is specific to each technical platform used to measure the expression of the genes or subsets of genes used in the probit regression model. The computer program calculates a score or probability and compares the score to a threshold. The sensitivity, specificity, and overall accuracy of each classifier is optimized by varying the threshold of classification using a receiver operating characteristic (ROC) curve.

[0303] A preliminary list of genes based on signatures from the Affymetrix platform (Affy signatures) as well as genes for the TLDA platform from other sources is provided below in Table 1A. The weights appropriate for the TLDA platform for the respective classifiers are then determined as described below in Example 4.

[0304] Table 1A: Preliminary list of genes used to form classifiers for the TLDA platform.

[0305]

[0306]

[0307]

[0308]

[0309]

[0310]

[0311]

[0312] Example 4: Bacterial / viral / SIRS classification using gene expression measured by RT-qPCR implemented on TLDA platform ​

[0313] The genes that constitute the three signature genes of the Host Response-ARI (HR-ARI) test were transformed into custom TaqMan® Low Density Array Cards from ThermoFisher Scientific (Waltham, MA). Custom multi-analyte quantitative real-time PCR (RT-qPCR) assays were used to measure the expression of these gene signature on the 384-well TaqMan Low Density Array (TLDA; Thermo-Fisher) platform. The TLDA cards were designed and manufactured with one or more TaqMan primer / probe sets per well, each representing a specific RNA transcript in the ARI signature, while multiple endogenous control RNA targets (TRAP1, PPIB, GAPDH, FPGS, DECR1, and 18S) were designed and manufactured for normalization to RNA loading and control for inter-plate variability. In practice, two reference genes (from five available) were selected that had the smallest coefficient of variation between all samples for the normalization process, and primer / probe sets with greater than 33% missing values (below the limit of quantification) were discarded. The remaining missing values, if any, were set to 1 + the maximum value (Cq), where Cq is the quantification cycle of the RT-qPCR. Normalized expression values were then calculated as the selected reference minus the average of the observed Cq values for any given primer / probe set. See Hellemans et al. (2007) Genome Biol 2007; 8(2): R19.

[0314] For each sample, a total of 174 unique primer / probe sets were assayed. Of these primers / probes, 144 primer / probe sets measured the gene targets of Affymetrix (microarray) probes representing 132 of the previously described three ARI gene signatures (i.e., genes in the bacterial gene expression signature, viral gene expression signature, and non-infectious gene expression signature); 6 probe sets were used for reference genes, and we additionally assayed 24 probe sets from the previously discovered pan-viral gene signature. See U.S. Patent No. 8,821,876; Zaas et al. Cell Host Microbe (2009) 6(3):207-217. In addition, multiple primer / probe sets for “replacement” genes were added for training, whose expression correlates with the expression of some genes from the Affymetrix signature. Some genes were replaced because the RT-qPCR assays for these genes performed poorly when performed using TLDA probes.

[0315] For each sample, total RNA was purified from PAXgene Blood RNA Tubes (PreAnalytix) and reverse transcribed to cDNA using Superscript VILO cDNA Synthesis Kit (Thermo-Fisher) according to the manufacturer's recommended procedure. A standard amount of cDNA for each sample was loaded into each master well and partitioned into each TaqMan assay well via microfluidic channels by centrifugation. TaqMan hydrolysis probes rely on 5' to 3' exonuclease activity to cleave the dual-labeled probe during hybridization to the complementary target sequence, and each round of amplification results in the production of a fluorescent signal. Quantitative detection of fluorescence "in real-time" indicates the accumulation of PCR product. The number of PCR cycles over the threshold of fluorescence signal during the exponential amplification and detection phase is the threshold cycle (C t ) or quantification cycle (C q ) as determined by commercial software for RT-qPCR instruments.

[0316] Sample / Group Selection:

[0317] Under IRB-approved procedures, we enrolled patients presenting to the emergency department with acute respiratory illness (see Table 11 below). Patients in this cohort were a subset of those reported in Table 1 of Tsalik et al. (2016) Sci Transl Med 9(322): 1-9, which is incorporated herein by reference. Retrospective clinical adjudication of clinical and other test data for these patients resulted in one of three assignments: bacterial ARI, viral ARI, or non-infectious illness.

[0318] Table 11: Demographic information for enrolled cohorts

[0319]

[0320] a Only subjects with viral, bacterial, or non-infectious illness were included in each validation cohort (when available).

[0321] b When mean age was not available or could not be calculated, data is given as adult or child.

[0322] c Non-infectious illness was defined by the presence of SIRS criteria, which includes at least two of the following four features: temperature < 36°C or > 38°C; heart rate > 90 beats per minute; respiratory rate 20 breaths per minute or arterial partial pressure of C02 < 32 mm Hg; and white blood cell count < 4000 or > 12,000 cells / mm 3 or > 10% band neutrophils.

[0323] d Three subjects did not report race.

[0324] M, male. F, female. B, black. W, white. O, other / unknown. GSE numbers refer to NCBI Gene Expression Omnibus datasets. N / A, not available based on public data.

[0325] Data analysis methods:

[0326] During the data pre-processing phase, we selected a subset of at least two reference gene targets (from the five available) that had the smallest coefficient of variation across all samples and plates. We discarded targets with more than 33% missing values (17 targets below the limit of quantification) (only in cases where these values were over-represented in any particular category such as bacterial ARI). Next we counted the remaining missing values as 1 + maximum value (C q ), and then normalized the expression values for all targets using the previously selected reference combination. In particular, we computed the normalized expression value as the average of the C q values for the selected references (DECR1 and PPIB) minus the average of the C

[0327] Once the data was normalized, we proceeded to build a classification model by fitting a sparse logistic regression model to the data (Friedman et al. (2010) J. Stat. Softw. 33, 1-22). This model estimates the probability that a subject belongs to a particular category as a weighted sum of the normalized gene targets. Specifically, we record that p(subject belongs to a category) = σ(w1x1 +... + w p x p ), where σ is the logistic function, w1,..., w p are the classification weights estimated during the fitting process, and x1,..., x p represent the p gene targets containing the normalized expression values.

[0328] Similar to the array-based classifiers, we constructed three binary classifiers: (1) bacterial ARI versus viral ARI and non-infectious disease; (2) viral ARI versus bacterial ARI and non-infectious disease; and (3) non-infectious disease versus bacterial and viral ARI. After fitting the three classifiers, we estimated p(bacterial ARI), p(viral ARI), and p(non-infectious disease). The threshold for each of the classifiers was chosen from the receiver operating characteristic (ROC) curve (expected sensitivity and specificity approximately equal) using a symmetric cost function (Fawcett (2006) Pattern Recogn Lett 27:861-874). As a result, if p(bacterial ARI) > t b , the subject was predicted to have a bacterial ARI, where t b is the threshold for the bacterial ARI classifier. Similarly, we selected the thresholds for the viral ARI and non-infectious disease classifiers, t v and t n , respectively. If desired, a combined prediction can be made by selecting the most likely case, i.e., the prediction with the largest probability, specifically, we recorded, argmax{p(bacterial ARI), p(viral ARI), p(non-infectious disease)}.

[0329] Results:

[0330] During the initial transition of the microarray-discovered genomic classifiers to the TLDA platform, we assayed 32 samples that were also assayed by microarray. This set of TLDA-based RT-qPCR measurements of the gene transcripts that constitute the ARI classifiers summarized the results obtained for the microarray-based measurements of the gene transcripts and, thus, was an effective method for classifying patients as having a bacterial or viral ARI or as having a non-infectious disease. We found that of the 32 samples tested on both the TLDA and microarray platforms, there was 84.4% agreement when assessed using their respective classifiers, meaning that 27 of the 32 subjects had the same combined prediction in the classification models based on both microarray and TLDA.

[0331] After confirming the agreement between microarray and TLDA-based classifications, we tested an additional 63 samples using TLDA-based classification in patients with a clinical diagnosis of ARI status but without previously characterized gene expression patterns. Thus, a total of 95 samples were evaluated using TLDA-based classification testing. This dataset from 95 samples allowed us to evaluate how the TLDA-based RT-qPCR platform would classify new patients using only the clinical diagnosis as a reference standard. In this experiment, we observed an overall accuracy of 81.1% which corresponds to 77 / 95 correctly classified samples. More specifically, the model achieved 80% (24 out of 30 correct), 77.4% (24 out of 31 correct), and 85.3% (29 out of 34 correct) accuracy for bacterial ARI, viral ARI, and non-infectious disease, respectively. In terms of performance of the individual classifiers, we observed an area under the ROC curve of 0.92, 0.86, and 0.91 for the bacterial ARI, viral ARI, and non-infectious disease classifiers, respectively. Provided that we do not count the validation dataset used for any of the classifiers, but we want an unbiased estimate of the classification performance (accuracy and area under the ROC curve), we report the performance metrics of leave-one-out cross-validation.

[0332] The weights and thresholds for each of the classifiers (bacterial ARI, viral ARI, and non-infectious disease) are shown in Table 12 below. Note that this table lists 151 gene targets instead of 174 gene targets because the reference genes were removed during the preprocessing stage, as well as the 17 targets that had missing values, as described above. These 17 targets were also removed during the preprocessing stage.

[0333] If the pan-viral signature genes were removed, we saw a slight decrease in performance, not more than 5% in all AUC, accuracy, and percentage of agreement values.

[0334] Summary:

[0335] The composite host response ARI classifier consists of gene expression signatures that identify bacterial ARI versus viral ARI, bacterial ARI versus non-infectious disease, and a mathematical classification framework. The mathematical classifier provides three discrete probabilities: that the subject has a bacterial ARI, a viral ARI, or a non-infectious disease. In each case, a cutoff or threshold value can be specified above which it will be determined that the patient has that condition. In addition, the threshold value can be modified to change the sensitivity and specificity of the test.

[0336] Measurement of these gene expression levels can be performed on multiple technical platforms. Here, we describe the measurement of these identifying markers using the TLDA-based RT-qPCR platform. In addition, the mathematical framework to determine the ARI etiology probability is adapted to this platform due to platform-specific training to accommodate the method of transcript measurement (i.e. establishing platform-specific weights, w1,..., w p A similar, straightforward approach can be performed to transfer the gene identifying markers to other gene expression detection platforms, and then train the relevant classifier. This example also demonstrates good agreement between the TLDA- and microarray-based classification of the etiology of ARIs. Finally, we show the use of the TLDA-based RT-qPCR platform and the associated mathematical classifier for the diagnosis of new patients with acute respiratory diseases.

[0337]

[0338]

[0339]

[0340]

[0341]

[0342]

[0343]

[0344] Any patent or publication mentioned in this specification is indicative of the levels of those skilled in the art to which the application pertains. These patents and publications are herein incorporated by reference to the same extent as if each individual publication was specifically and individually incorporated by reference. In case of conflict, the present specification, including that of definitions, will control.

[0345] Those skilled in the art will readily understand that the application is well adapted to carry out the objects and obtain the results and advantages mentioned, as well as others inherent therein. The present application described herein is presently representative of preferred embodiments, is exemplary, and is not intended as limitations on the scope of the application. Changes therein and other uses will suggest themselves to those skilled in the art upon a reading of the foregoing description. Such changes and other uses are considered to be within the spirit of the application as defined by the scope of the claims.

Claims

1. A method for manufacturing acute respiratory disease classifiers for a platform, wherein the classifiers comprise a bacterial ARI classifier, a viral ARI classifier, and a non-infectious disease classifier for the platform, the method comprising: (a) obtaining biological samples from a plurality of subjects known to have a bacterial acute respiratory infection; (b) obtaining biological samples from a plurality of subjects known to have a viral acute respiratory infection; (c) obtaining biological samples from a plurality of subjects known to have a non-infectious disease; (d) measuring gene expression levels of a plurality of genes (e.g., all expressed genes or transcriptomes, or a subset thereof) in each of the biological samples from steps (a), (b), and (c) on the platform; (e) normalizing the gene expression levels resulting in step (d) to generate normalized gene expression values; and (f) generating a bacterial ARI classifier, a viral ARI classifier, and a non-infectious disease classifier for the platform based on the normalized gene expression values, thereby manufacturing the acute respiratory disease classifiers for the platform.

2. The method of claim 1, wherein the platform comprises an array platform, a thermal cycler platform (e.g., multiplex and / or real-time PCR platform), a hybridization and multi-signal encoded (e.g., fluorescent) detector platform, a nucleic acid mass spectrometry platform, a nucleic acid sequencing platform, or a combination thereof.

3. The method of claim 1, wherein the generating iteratively comprises: (i) assigning a weight to each normalized gene expression value, inputting the weight and expression value for each gene into a classifier (e.g., linear regression classifier) equation and determining a score for the outcome for each of the plurality of subjects, and then (ii) determining the accuracy of the classification of each outcome among the plurality of subjects, and then (iii) adjusting the weights until the accuracy of the classification is optimized, to provide the bacterial ARI classifier, viral ARI classifier, and non-infectious disease classifier for the platform, wherein the genes with non-zero weights are included in the respective classifiers, and optionally uploading the elements (genes, weights, and / or etiology thresholds) of each classifier to one or more databases.

4. A bacterial ARI classifier manufactured according to the method of any one of claims 1-3, wherein the bacterial ARI classifier comprises expression levels of 5, 10, 20, 30, or 50, to 80, 100, 150, or 200 of the genes (measurable, e.g., having oligonucleotide probes homologous to the genes) listed as part of the viral ARI classifier in Table 1, Table 2, Table 9, Table 10, and / or Table 12.

5. A viral ARI classifier manufactured according to the method of any one of claims 1-3, wherein the viral classifier comprises the expression level of 5, 10, 20, 30, or 50, to 80, 100, 150, or 200 of the genes listed as part of a viral ARI classifier in Table 1, Table 2, Table 9, Table 10, and / or Table 12 (measurable, e.g., having an oligonucleotide probe homologous to the gene).

6. A non-infectious disease classifier manufactured according to the method of any one of claims 1-3, the non-infectious classifier comprising the expression level of 5, 10, 20, 30, or 50, to 80, 100, 150, or 200 of the genes listed as part of a non-infectious disease classifier in Table 1, Table 2, Table 9, Table 10, and / or Table 12 (measurable, e.g., having an oligonucleotide probe homologous to the gene).

7. A method for determining the etiology of an acute respiratory disease in a subject having or at risk of having the acute respiratory disease, the acute respiratory disease selected from bacterial, viral, and / or non-infectious, the method comprising: (a) obtaining a biological sample from the subject; (b) measuring gene expression levels of a predefined set of genes (i.e., identifying markers) in the biological sample on a platform; (c) normalizing the gene expression levels to generate normalized gene expression values; (d) inputting the normalized gene expression values into one or more acute respiratory disease classifiers selected from a bacterial acute respiratory infection (ARI) classifier, a viral ARI classifier, and a non-infectious disease classifier, the one or more classifiers comprising a predefined weighting value (i.e., coefficient) for each of the genes of the predefined set of genes for the platform, optionally, wherein the one or more classifiers are obtained from one or more databases; and (e) calculating a probability of etiology for one or more of a bacterial ARI, a viral ARI, and a non-infectious disease based on the normalized gene expression values and the one or more classifiers, thereby determining whether the acute respiratory disease in the subject is of bacterial origin, viral origin, non-infectious origin, or some combination thereof.

8. A method of treating an acute respiratory disease in a subject, the method comprising administering an appropriate treatment regimen to the subject based on the etiology determined by the method of claim 7.

9. A method of monitoring the response of a vaccine or drug in a subject having or at risk of having an acute respiratory disease selected from bacterial, viral, and / or non-infectious, the method comprising determining the host response of the subject by the method of claim 7.

10. A system for determining the etiology of an acute respiratory disease selected from bacterial, viral, and / or non-infectious in a subject, the system comprising: at least one processor; a sample input loop configured to receive a biological sample from the subject; a sample analysis circuit coupled with the at least one processor and configured to determine gene expression levels of the biological sample; an input / output circuit coupled with the at least one processor; a storage circuit coupled with the at least one processor and configured to store data, parameters, and / or classifiers; and a memory coupled with the processor and comprising computer readable program code embodied in the memory, which when executed by the at least one processor, causes at least one processor to perform operations comprising: controlling / conducting, by the sample analysis circuit, measurement of gene expression levels of a predefined set of genes (i.e., identifying markers) in the biological sample; normalizing the gene expression levels to generate normalized gene expression values; obtaining, from the storage circuit, a bacterial Acute Respiratory Infection (ARI) classifier, a viral ARI classifier, and a non-infectious disease classifier, the one or more classifiers comprising predefined weighted values (i.e., coefficients) for each of the genes of the predefined set of genes; inputting the normalized gene expression values into one or more Acute Respiratory Disease classifiers selected from the bacterial Acute Respiratory Infection (ARI) classifier, the viral ARI classifier, and the non-infectious disease classifier; calculating, based on the one or more classifiers, a probability of etiology for one or more of a bacterial ARI, a viral ARI, and a non-infectious disease; and controlling output, by the input / output circuit, to determine whether the Acute Respiratory Disease in the subject is of bacterial origin, viral origin, non-infectious origin, or some combination thereof.

Images